Air conditioner heat recovery intelligent management system and method based on multi-source data fusion

Abstract

The invention discloses an air conditioner heat recovery intelligent management system and method based on multi-source data fusion, and relates to the technical field of big data analysis. Constructing a heat recovery characteristic set associated with the process exhaust state and the heat recovery performance; a fluctuation difference value is extracted through time sequence analysis, a standardized working condition feature portrait is constructed, three types of working conditions are analyzed and divided in combination with a sliding window, and production normal fluctuation and abnormal fluctuation precursor are accurately distinguished; and training an exclusive neural network early warning model based on the three types of working condition sets, constructing working condition feature vectors in real time, matching the model to predict power fluctuation, and sending out abnormal early warning in time. The problems of regulation lag, insufficient adaptability and fuzzy fluctuation judgment are solved, the system operation stability and the energy utilization efficiency are effectively improved, and the production continuity and the product quality stability of a chemical fiber factory are guaranteed.

Description

Technical Field

[0001] This invention relates to the field of big data analytics, specifically to an intelligent management system and method for air conditioning heat recovery based on multi-source data fusion. Background Technology

[0002] As an important pillar industry of the national economy, chemical fiber plants undertake the mission of supplying basic materials for multiple fields such as textiles and building materials. Their stable and efficient production is crucial to the coordinated development of the industrial chain. The chemical fiber manufacturing process involves multiple thermochemical reactions and requires a continuous input of a large amount of heat energy. The stable supply and efficient utilization of heat energy are directly related to the continuity of production, the stability of product quality, and the level of energy consumption control.

[0003] Existing air conditioning heat recovery systems in chemical fiber plants mostly rely on monitoring a single parameter at the heat recovery unit level for regulation, lacking direct detection and systematic analysis of characteristic parameters of process exhaust air. Furthermore, small fluctuations in exhaust air parameters are prone to occur during the manufacturing process, and current technologies lack effective methods to determine the rationality of such fluctuations. This makes it difficult to distinguish between reasonable fluctuations from normal production and precursors to abnormal fluctuations, and also fails to provide accurate data support for heat recovery regulation. This results in lag in regulation and insufficient adaptability, ultimately affecting the operational stability and energy efficiency of the heat recovery system. Therefore, there is an urgent need for an intelligent management system and method for air conditioning heat recovery based on multi-source data fusion. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent management system and method for air conditioning heat recovery based on multi-source data fusion, so as to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent management method for air conditioning heat recovery based on multi-source data fusion, the intelligent management method for air conditioning heat recovery comprising the following steps: Step S1: According to the preset data collection period, collect data on the process exhaust generated by the manufacturing process of the chemical fiber plant, obtain the process exhaust characteristic parameters, and construct a process exhaust characteristic set based on the data collection timestamps of each data collection period and the process exhaust characteristic parameters. Step S1-1: By deploying temperature sensors, humidity sensors, air volume sensors, and static pressure sensors on the process exhaust ducts of the chemical fiber plant's manufacturing process, the characteristic parameters of the process exhaust are continuously collected according to a preset data acquisition cycle; the characteristic parameters of the process exhaust include the temperature, humidity, air volume, and static pressure of the process exhaust; the static pressure refers to the static pressure of the process exhaust within the process exhaust ducts of the chemical fiber plant's manufacturing process. Step S1-2: Associate and bind each data acquisition timestamp with the process exhaust characteristic parameters collected at the corresponding time. Select the data acquisition timestamp as the key and the process exhaust characteristic parameter corresponding to the data acquisition timestamp as the value, and store them in a structured manner to form a process exhaust characteristic set. By deploying temperature sensors, humidity sensors, air volume sensors, and static pressure sensors in the process exhaust ducts of the chemical fiber plant, continuous and accurate collection of core characteristic parameters of process exhaust temperature, humidity, air volume, and static pressure is achieved according to a preset data acquisition cycle, ensuring the real-time and completeness of data acquisition. Using the data acquisition timestamp as the key, the process exhaust characteristic parameters at the corresponding time are associated, bound, and stored in a structured manner to form a process exhaust characteristic set, providing standardized and traceable basic data support for subsequent multi-source data fusion analysis and fluctuation feature extraction.

[0006] Step S2: Synchronously collect heat exchange parameters from the heat recovery device side based on the data acquisition cycle to obtain instantaneous heat recovery power; read the process exhaust characteristic parameters in the corresponding process exhaust characteristic set according to the data acquisition timestamp as the associated index; after normalizing each process exhaust characteristic parameter, construct a heat recovery characteristic set in key-value pair storage form by combining the instantaneous heat recovery power. Step S2-1: Based on a preset data acquisition cycle, synchronously acquire heat exchange parameters from the heat recovery device side. The heat exchange parameters include the inlet temperature of the recovery medium, the outlet temperature of the recovery medium, and the flow rate of the recovery medium. The flow rate of the recovery medium is acquired by deploying a flow meter sensor. The instantaneous heat recovery power is calculated based on the acquired heat exchange parameters. The formula for calculating the instantaneous heat recovery power is as follows: P = c × q × (T) out -T in ); In the formula, P represents the instantaneous heat recovery power, which is the quantified value of the heat recovery capacity of the heat recovery device at a single data acquisition time stamp; c represents the specific heat capacity of the recovery medium, which is a physical characteristic parameter of the recovery medium itself, i.e., the specific heat capacity of the real-time process exhaust air; q represents the flow rate of the recovery medium, which is collected by a flow meter sensor deployed on the side of the heat recovery device; T out T represents the outlet temperature of the recovered medium, which is the temperature parameter when the recovered medium flows out from the heat recovery unit; in This is expressed as the inlet temperature of the recovery medium, which is the temperature parameter when the recovery medium flows into the heat recovery device. Step S2-2: Select the data acquisition timestamp as the associated index, and read the process exhaust feature parameters corresponding to the data acquisition timestamp from the process exhaust feature set; retrieve the historical maximum and minimum values ​​corresponding to each sub-parameter in the process exhaust feature parameters, and perform normalization processing on each sub-parameter in the currently acquired process exhaust feature parameters. The normalization processing process is to subtract the corresponding historical minimum value from the current sub-parameter, and then divide by the difference between the corresponding historical maximum value and the historical minimum value. Step S2-3: Associate and bind the normalized process exhaust characteristic parameters with the instantaneous heat recovery power, select the data acquisition timestamp as the key, and select the parameter group composed of the normalized process exhaust characteristic parameters and the instantaneous heat recovery power as the value, and store it in a structured way to form a heat recovery feature set in key-value pair storage format. Based on a preset data acquisition cycle, the inlet temperature, outlet temperature, and flow rate parameters of the recovery medium on the heat recovery device side are synchronously collected to accurately calculate the instantaneous heat recovery power, realizing the quantitative acquisition of core performance data of heat recovery. Using the data acquisition timestamp as the correlation index, the corresponding process exhaust characteristic parameters are retrieved and normalized to eliminate the dimensional differences between different parameters. Then, the normalized exhaust parameters are correlated and bound with the instantaneous heat recovery power to construct a key-value pair form of heat recovery characteristic set. This successfully establishes a direct correlation between process exhaust parameters and heat recovery power, providing a fused and homogeneous data foundation for subsequent fluctuation correlation analysis and solving the control problem of isolated data and lack of correlation basis.

[0007] Step S3: Obtain the historical heat recovery feature set for data analysis to obtain the instantaneous heat recovery power fluctuation difference and the heat recovery feature fluctuation difference set between two adjacent data acquisition timestamps. Based on the heat recovery feature fluctuation difference set, analyze and extract features to construct the working condition feature profile corresponding to the instantaneous heat recovery power fluctuation difference. Step S3-1: Sort the historical heat recovery feature set according to the order of data acquisition timestamps, extract the instantaneous heat recovery power corresponding to two adjacent data acquisition timestamps after sorting, calculate the difference between the instantaneous heat recovery power of the later data acquisition timestamp and the instantaneous heat recovery power of the earlier data acquisition timestamp, and record it as the instantaneous heat recovery power fluctuation difference. Step S3-2: Extract each sub-parameter from the normalized process exhaust characteristic parameters corresponding to two adjacent data acquisition timestamps after sorting, including temperature, humidity, air volume, and static pressure. Calculate the difference between the sub-parameters of the later data acquisition timestamp and the sub-parameters of the earlier data acquisition timestamp. Associate the temperature difference, humidity difference, air volume difference, and static pressure difference with the corresponding instantaneous heat recovery power fluctuation difference to form a heat recovery characteristic fluctuation difference set. The key of the heat recovery characteristic fluctuation difference set is the instantaneous heat recovery power fluctuation difference, and the value is a quadruple parameter group containing the temperature difference, humidity difference, air volume difference, and static pressure difference. Step S3-3: Perform structured analysis on each quaternary parameter group in the heat recovery characteristic fluctuation difference set to establish temperature difference, humidity difference, air volume difference, and static pressure difference as four independent core feature dimensions; assign a fixed feature identifier to each core feature dimension, where temperature difference corresponds to temperature fluctuation feature, humidity difference corresponds to humidity fluctuation feature, air volume difference corresponds to air volume fluctuation feature, and static pressure difference corresponds to static pressure fluctuation feature. Step S3-4: Map and associate the specific difference data corresponding to each core feature dimension with the feature identifier, and integrate them into a standardized data structure in a fixed order of temperature fluctuation feature, humidity fluctuation feature, air volume fluctuation feature, and static pressure fluctuation feature. This standardized data structure is the working condition feature profile corresponding to the instantaneous heat recovery power fluctuation difference. The historical heat recovery feature set is sorted chronologically by timestamp, and the instantaneous heat recovery power fluctuation difference and the fluctuation difference of each sub-parameter of the process exhaust are calculated for adjacent timestamps. A heat recovery feature fluctuation difference set is constructed to realize the quantitative extraction of fluctuation data. By analyzing the fluctuation difference set, four core feature dimensions—temperature difference, humidity difference, air volume difference, and static pressure difference—are established and assigned fixed labels. They are then integrated in a fixed order to form an operating condition feature profile, transforming the scattered fluctuation data into a structured and concrete feature carrier. This clearly presents the correspondence between exhaust parameter fluctuations and heat recovery power fluctuations, providing a standardized feature basis for the rationality judgment of fluctuations and solving the problem of fuzzy and inaccurate fluctuation features.

[0008] Step S4: Set up a sliding window for working condition feature profiles to extract and analyze data from multiple working condition feature profiles. Calculate the data division benchmark value based on the difference in instantaneous heat recovery power fluctuations extracted from multiple profiles. Divide the extracted working condition feature profiles into data based on the data division benchmark value to obtain the first working condition feature profile set, the second working condition feature profile set, and the third working condition feature profile set. Step S4-1: Based on the window length of the preset working condition feature profile sliding window, and using the number of data acquisition timestamps as the unit of measurement, slide and cut the continuous historical working condition feature profiles according to the order of the data acquisition timestamps in the historical heat recovery feature set. After each sliding cut, the window moves backward at the interval of a single data acquisition timestamp until the cutting operation of all working condition feature profiles is completed. Step S4-2: Extract all instantaneous heat recovery power fluctuation differences contained in the sliding window of each captured working condition feature profile, calculate the extracted instantaneous heat recovery power fluctuation differences to obtain the corresponding arithmetic mean, and calculate the standard deviation of the set of instantaneous heat recovery power fluctuation differences. Step S4-3: Analyze and process all the instantaneous heat recovery power fluctuation differences according to the three sigma principle, and select the calculated arithmetic mean as the center value. Subtract three times the standard deviation from the arithmetic mean to obtain the lower limit of fluctuation, and add three times the standard deviation to the arithmetic mean to obtain the upper limit of fluctuation. Based on the lower limit and the upper limit of fluctuation, the normal fluctuation range of instantaneous heat recovery power is formed. Step S4-4: Compare the instantaneous heat recovery power fluctuation difference corresponding to each operating condition feature portrait in the sliding window with the normal fluctuation range and the center value. The operating condition feature portraits whose instantaneous heat recovery power fluctuation difference equals the center value are assigned to the first operating condition feature portrait set; the operating condition feature portraits whose instantaneous heat recovery power fluctuation difference is greater than the lower fluctuation limit and less than the center value are assigned to the second operating condition feature portrait set; and the operating condition feature portraits whose instantaneous heat recovery power fluctuation difference is greater than the center value and less than the upper fluctuation limit are assigned to the third operating condition feature portrait set. The index of each operating condition feature portrait set is the instantaneous heat recovery power fluctuation difference, and the value is the operating condition feature portrait corresponding to the instantaneous heat recovery power fluctuation difference. By using a preset sliding window to capture and dynamically analyze the operating condition feature profile over time, the arithmetic mean and standard deviation of the instantaneous heat recovery power fluctuation difference within the window are calculated. Combined with the three sigma principle, the normal fluctuation range and center value, lower limit of fluctuation, and upper limit of fluctuation of instantaneous heat recovery power are established as data classification benchmarks. Based on the benchmark values, the operating condition feature profile is divided into three independent sets, realizing accurate classification of different fluctuating operating conditions, clarifying the boundaries and levels of reasonable fluctuations, and providing clear operating condition classification basis for targeted regulation. This solves the problems of ambiguous fluctuation classification and inability to adapt to different operating conditions.

[0009] Step S5: Train the data of each working condition feature profile set through the neural network model to obtain the instantaneous heat recovery power fluctuation early warning model corresponding to the data division benchmark value. Collect real-time data of the process exhaust of the core process link of the chemical fiber plant, analyze and process it synchronously to obtain the real-time working condition feature profile, divide it based on the data division benchmark value, and provide early warning prompts for air conditioning heat recovery through the corresponding instantaneous heat recovery power fluctuation early warning model. Step S5-1: Take the first set of working condition feature profiles, the second set of working condition feature profiles, and the third set of working condition feature profiles as independent training datasets; perform feature vector transformation on the working condition feature profiles in each dataset, extract the temperature fluctuation feature data, humidity fluctuation feature data, air volume fluctuation feature data, and static pressure fluctuation feature data from each working condition feature profile, and combine them in a fixed order of temperature fluctuation feature, humidity fluctuation feature, air volume fluctuation feature, and static pressure fluctuation feature to construct a four-dimensional working condition feature vector; calculate the Euclidean modulus of each four-dimensional working condition feature vector, use the Euclidean modulus as the independent variable, and use the instantaneous heat recovery power fluctuation difference associated with the corresponding working condition feature profile as the dependent variable, and input them into three independent neural network models for training. After training, the first instantaneous heat recovery power fluctuation early warning model, the second instantaneous heat recovery power fluctuation early warning model, and the third instantaneous heat recovery power fluctuation early warning model are obtained. The formula for calculating the European modulus is as follows: ; In the formula, L represents the Euclidean modulus of the four-dimensional operating condition feature vector, which is a quantitative comprehensive index of the operating condition features; △T represents the temperature fluctuation difference, i.e., the temperature fluctuation feature data in the operating condition feature profile; △H represents the humidity fluctuation difference, i.e., the humidity fluctuation feature data in the operating condition feature profile; △Q represents the air volume fluctuation difference, i.e., the air volume fluctuation feature data in the operating condition feature profile; and △S represents the static pressure fluctuation difference, i.e., the static pressure fluctuation feature data in the operating condition feature profile. Step S5-2: Extract four feature data from the real-time working condition feature profile, combine them in the same fixed order to construct a real-time four-dimensional working condition feature vector, and calculate the real-time Euclidean modulus of the real-time four-dimensional working condition feature vector. Step S5-3: Compare the real-time instantaneous heat recovery power fluctuation difference corresponding to the real-time operating condition feature profile with the data division benchmark value, and match the corresponding instantaneous heat recovery power fluctuation early warning model. When the real-time instantaneous heat recovery power fluctuation difference is equal to the center value, the first early warning model is matched; when it is greater than the lower limit of the fluctuation but less than the center value, the second early warning model is matched; when it is greater than the center value but less than the upper limit of the fluctuation, the third early warning model is matched. Input the real-time Euclidean modulus into the matched instantaneous heat recovery power fluctuation early warning model, and output the instantaneous heat recovery power fluctuation prediction value corresponding to the first data acquisition timestamp in the next data acquisition cycle, which is recorded as the fluctuation prediction value. Compare the fluctuation prediction value with the normal fluctuation range of instantaneous heat recovery power. When the fluctuation prediction value is less than the lower limit of the fluctuation or greater than the upper limit of the fluctuation, an air conditioning heat recovery adjustment early warning prompt is issued; otherwise, no early warning prompt is triggered, and intelligent monitoring and management of air conditioning heat recovery continues. Three types of operating condition feature profiles are used as independent training datasets. The operating condition feature profiles are transformed into four-dimensional feature vectors and Euclidean modulus is calculated. The modulus is used as the independent variable and the corresponding fluctuation difference is used as the dependent variable to train three types of dedicated instantaneous heat recovery power fluctuation early warning models, achieving accurate model-to-operating condition adaptation. By collecting process exhaust and heat recovery device data in real time, real-time operating condition feature profiles and feature vectors are constructed. The corresponding early warning models are matched to predict the fluctuation value in the next period. After comparison with the normal fluctuation range, an early warning is triggered, realizing early prediction and accurate early warning of abnormal fluctuations in heat recovery power. This significantly improves the adaptability of the heat recovery system to small process fluctuations, while enhancing system operation stability and energy utilization efficiency, solving the problems of slow response and insufficient adaptability of traditional control.

[0010] Furthermore, an intelligent management system for air conditioning heat recovery based on multi-source data fusion is provided. This intelligent management system for air conditioning heat recovery includes an exhaust data acquisition module, a heat recovery feature integration module, an operating condition feature profile construction module, an operating condition set division module, and a heat recovery early warning and monitoring module. The exhaust data acquisition module is used to collect process exhaust characteristic parameters of the chemical fiber plant's manufacturing process and construct a process exhaust characteristic set; the heat recovery characteristic integration module is used to collect heat exchange parameters from the heat recovery device and combine them with normalized process exhaust characteristic parameters to construct a heat recovery characteristic set; the operating condition characteristic profile construction module is used to analyze historical heat recovery characteristic sets to obtain fluctuation differences and construct operating condition characteristic profiles based on the fluctuation difference set; the operating condition set division module is used to extract operating condition characteristic profiles through a sliding window, calculate a benchmark value based on the instantaneous heat recovery power fluctuation difference, and divide the operating condition characteristic profile set; the heat recovery early warning monitoring module is used to train an instantaneous heat recovery power fluctuation early warning model, predict based on real-time data, and trigger air conditioning heat recovery early warnings. The exhaust data acquisition module includes a parameter acquisition unit and a feature set construction unit. The parameter acquisition unit is used to continuously acquire the temperature, humidity, air volume and static pressure of the process exhaust air through deployed sensors according to a preset data acquisition cycle. The feature set construction unit is used to associate and bind each data acquisition timestamp with the corresponding process exhaust air feature parameter, and store them in a structured manner to form a process exhaust air feature set. The heat recovery feature integration module includes a heat exchange parameter acquisition unit and a feature set integration unit. The heat exchange parameter acquisition unit is used to simultaneously acquire the inlet temperature, outlet temperature, and flow rate of the recovery medium and calculate the instantaneous heat recovery power. The feature set integration unit is used to normalize the process exhaust characteristic parameters, then associate and bind them with the instantaneous heat recovery power, and store them in a structured manner to form a heat recovery feature set. The operating condition feature profile construction module includes a fluctuation difference calculation unit and a feature profile generation unit. The fluctuation difference calculation unit is used to calculate the instantaneous heat recovery power fluctuation difference and the difference of each sub-parameter of the process exhaust at adjacent data acquisition timestamps and associate them to form a heat recovery feature fluctuation difference set. The feature profile generation unit is used to parse the heat recovery feature fluctuation difference set, establish core feature dimensions and assign feature identifiers, and integrate them in a fixed order to form an operating condition feature profile. The operating condition set division module includes a sliding window interception unit and a feature set division unit; the sliding window interception unit is used to slide and intercept continuous historical operating condition feature profiles according to a preset window length and timestamp order; the feature set division unit is used to calculate the arithmetic mean and standard deviation of the instantaneous heat recovery power fluctuation difference, establish the normal fluctuation range, and divide the operating condition feature profiles into three sets. The heat recovery early warning and monitoring module includes an early warning model training unit and a real-time early warning and monitoring unit. The early warning model training unit uses three sets of operating condition feature profiles as training datasets, and trains three corresponding instantaneous heat recovery power fluctuation early warning models through feature vector transformation and Euclidean modulus calculation. The real-time early warning and monitoring unit collects real-time process exhaust and heat exchange parameters to construct real-time operating condition feature profiles, matches the predicted fluctuation values ​​of the corresponding early warning models, and determines whether to trigger an early warning after comparing them with the normal fluctuation range.

[0011] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention uses a multi-source data collaborative acquisition and fusion processing mechanism to simultaneously acquire core parameters of the process exhaust air temperature, humidity, air volume, and static pressure of the chemical fiber plant, as well as heat exchange parameters from the heat recovery device. After normalization processing to eliminate dimensional differences, a set of correlated heat recovery features is constructed, breaking the data isolation barrier of traditional single-parameter monitoring. This enables a direct correlation between the process exhaust air status and heat recovery performance, providing comprehensive and accurate data support for heat recovery regulation and ensuring the integrity and adaptability of the data.

[0012] 2. This invention utilizes a structured analysis of fluctuation characteristics and a precise classification method for operating conditions. By extracting the power fluctuation difference and exhaust parameter fluctuation difference between adjacent acquisition cycles through time-series analysis, it constructs a standardized operating condition feature profile. Combining a sliding window and the three sigma principle, it establishes the normal fluctuation range and accurately divides three types of operating condition sets. This effectively distinguishes between normal production fluctuations and abnormal fluctuation precursors, solving the problem of ambiguous fluctuation identification in traditional technologies. It provides clear operating condition boundaries and hierarchical classification criteria for targeted regulation, improving the accuracy and reliability of fluctuation identification.

[0013] 3. This invention employs a working condition-adaptive neural network early warning model and a real-time control mechanism. Based on three types of working conditions, a dedicated early warning model is trained. By constructing and matching real-time working condition feature vectors with the model, the fluctuation value of heat recovery power in the next period can be predicted in advance. This enables accurate prediction and timely early warning of abnormal fluctuations, significantly improving the problems of traditional control lag and insufficient adaptability. It greatly enhances the operational stability of the air conditioning heat recovery system, while optimizing energy utilization efficiency and ensuring the production continuity and product quality stability of the chemical fiber plant. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating an intelligent management method for air conditioning heat recovery based on multi-source data fusion according to the present invention. Figure 2 This is a schematic diagram of the structure of an intelligent management system for air conditioning heat recovery based on multi-source data fusion according to the present invention. Detailed Implementation

[0015] 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.

[0016] Example 1: As Figure 1 As shown, this invention provides a technical solution: an intelligent management method for air conditioning heat recovery based on multi-source data fusion. The intelligent management method for air conditioning heat recovery includes the following steps: Step S1: According to the preset data collection period, collect data on the process exhaust generated by the manufacturing process of the chemical fiber plant, obtain the process exhaust characteristic parameters, and construct a process exhaust characteristic set based on the data collection timestamps of each data collection period and the process exhaust characteristic parameters. Step S1-1: By deploying temperature sensors, humidity sensors, air volume sensors, and static pressure sensors on the process exhaust ducts of the chemical fiber plant's manufacturing process, the characteristic parameters of the process exhaust are continuously collected according to a preset data acquisition cycle; the characteristic parameters of the process exhaust include the temperature, humidity, air volume, and static pressure of the process exhaust; the static pressure refers to the static pressure of the process exhaust within the process exhaust ducts of the chemical fiber plant's manufacturing process. Step S1-2: Associate and bind each data acquisition timestamp with the process exhaust characteristic parameters collected at the corresponding time. Select the data acquisition timestamp as the key and the process exhaust characteristic parameter corresponding to the data acquisition timestamp as the value, and store them in a structured manner to form a process exhaust characteristic set. In practical implementation, temperature sensors, humidity sensors, air volume sensors, and static pressure sensors are deployed at key nodes of the exhaust ducts in the chemical fiber plant's manufacturing process to ensure that the data collection range covers the core exhaust path. By continuously collecting key characteristic parameters and storing them with timestamps as keys, data traceability and structured management are achieved. Sensors are calibrated regularly to avoid parameter distortion caused by equipment errors, ensuring that the data collection cycle is adapted to the process production rhythm and guaranteeing the integrity and timeliness of the feature set.

[0017] Step S2: Synchronously collect heat exchange parameters from the heat recovery device side based on the data acquisition cycle to obtain instantaneous heat recovery power; read the process exhaust characteristic parameters in the corresponding process exhaust characteristic set according to the data acquisition timestamp as the associated index; after normalizing each process exhaust characteristic parameter, construct a heat recovery characteristic set in key-value pair storage form by combining the instantaneous heat recovery power. Step S2-1: Based on a preset data acquisition cycle, synchronously acquire heat exchange parameters from the heat recovery device side. The heat exchange parameters include the inlet temperature of the recovery medium, the outlet temperature of the recovery medium, and the flow rate of the recovery medium. The flow rate of the recovery medium is acquired by deploying a flow meter sensor. The instantaneous heat recovery power is calculated based on the acquired heat exchange parameters. Step S2-2: Select the data acquisition timestamp as the associated index, and read the process exhaust feature parameters corresponding to the data acquisition timestamp from the process exhaust feature set; retrieve the historical maximum and minimum values ​​corresponding to each sub-parameter in the process exhaust feature parameters, and perform normalization processing on each sub-parameter in the currently acquired process exhaust feature parameters. The normalization processing process is to subtract the corresponding historical minimum value from the current sub-parameter, and then divide by the difference between the corresponding historical maximum value and the historical minimum value. Step S2-3: Associate and bind the normalized process exhaust characteristic parameters with the instantaneous heat recovery power, select the data acquisition timestamp as the key, and select the parameter group composed of the normalized process exhaust characteristic parameters and the instantaneous heat recovery power as the value, and store it in a structured way to form a heat recovery feature set in key-value pair storage format. In practical implementation, the timing of data acquisition by the sensors on the heat recovery device side and the process exhaust side should be kept consistent. Heat exchange parameters such as inlet temperature, outlet temperature, and flow rate of the recovery medium should be collected synchronously. Instantaneous heat recovery power is obtained by quantitative calculation. The corresponding process exhaust characteristic parameters are retrieved using the data acquisition timestamp as the associated index. Based on the extreme values ​​of historical data acquisition under all operating conditions, each sub-parameter is normalized to eliminate dimensional differences. The normalized parameters are associated and bound with the instantaneous heat recovery power and stored in a structured key-value pair format to form a heat recovery characteristic set. The timetamp correspondence is strictly verified to avoid data misalignment affecting the reliability of the analysis.

[0018] Step S3: Obtain the historical heat recovery feature set for data analysis to obtain the instantaneous heat recovery power fluctuation difference and the heat recovery feature fluctuation difference set between two adjacent data acquisition timestamps. Based on the heat recovery feature fluctuation difference set, analyze and extract features to construct the working condition feature profile corresponding to the instantaneous heat recovery power fluctuation difference. Step S3-1: Sort the historical heat recovery feature set according to the order of data acquisition timestamps, extract the instantaneous heat recovery power corresponding to two adjacent data acquisition timestamps after sorting, calculate the difference between the instantaneous heat recovery power of the later data acquisition timestamp and the instantaneous heat recovery power of the earlier data acquisition timestamp, and record it as the instantaneous heat recovery power fluctuation difference. Step S3-2: Extract each sub-parameter from the normalized process exhaust characteristic parameters corresponding to two adjacent data acquisition timestamps after sorting, including temperature, humidity, air volume, and static pressure. Calculate the difference between the sub-parameters of the later data acquisition timestamp and the sub-parameters of the earlier data acquisition timestamp. Associate the temperature difference, humidity difference, air volume difference, and static pressure difference with the corresponding instantaneous heat recovery power fluctuation difference to form a heat recovery characteristic fluctuation difference set. The key of the heat recovery characteristic fluctuation difference set is the instantaneous heat recovery power fluctuation difference, and the value is a quadruple parameter group containing the temperature difference, humidity difference, air volume difference, and static pressure difference. Step S3-3: Perform structured analysis on each quaternary parameter group in the heat recovery characteristic fluctuation difference set to establish temperature difference, humidity difference, air volume difference, and static pressure difference as four independent core feature dimensions; assign a fixed feature identifier to each core feature dimension, where temperature difference corresponds to temperature fluctuation feature, humidity difference corresponds to humidity fluctuation feature, air volume difference corresponds to air volume fluctuation feature, and static pressure difference corresponds to static pressure fluctuation feature. Step S3-4: Map and associate the specific difference data corresponding to each core feature dimension with the feature identifier, and integrate them into a standardized data structure in a fixed order of temperature fluctuation feature, humidity fluctuation feature, air volume fluctuation feature, and static pressure fluctuation feature. This standardized data structure is the working condition feature profile corresponding to the instantaneous heat recovery power fluctuation difference. In practical implementation, the historical heat recovery feature set is sorted chronologically according to the data acquisition timestamps. The instantaneous heat recovery power and normalized process exhaust sub-parameters corresponding to adjacent timestamps are extracted, and the fluctuation difference between the two is calculated. The temperature difference, humidity difference, air volume difference, static pressure difference and the corresponding power fluctuation difference are correlated to form a heat recovery feature fluctuation difference set. The four-element parameter group is analyzed to establish four core feature dimensions and assigned fixed labels. The parameters are mapped and integrated in a fixed order to form a standardized operating condition feature profile. The feature dimensions are defined in accordance with the process influence logic, and the parameter group structure is kept consistent to ensure that the profile truly reflects the correlation of operating condition fluctuations.

[0019] Step S4: Set up a sliding window for working condition feature profiles to extract and analyze data from multiple working condition feature profiles. Calculate the data division benchmark value based on the difference in instantaneous heat recovery power fluctuations extracted from multiple profiles. Divide the extracted working condition feature profiles into data based on the data division benchmark value to obtain the first working condition feature profile set, the second working condition feature profile set, and the third working condition feature profile set. Step S4-1: Based on the window length of the preset working condition feature profile sliding window, and using the number of data acquisition timestamps as the unit of measurement, slide and cut the continuous historical working condition feature profiles according to the order of the data acquisition timestamps in the historical heat recovery feature set. After each sliding cut, the window moves backward at the interval of a single data acquisition timestamp until the cutting operation of all working condition feature profiles is completed. Step S4-2: Extract all instantaneous heat recovery power fluctuation differences contained in the sliding window of each captured working condition feature profile, calculate the extracted instantaneous heat recovery power fluctuation differences to obtain the corresponding arithmetic mean, and calculate the standard deviation of the set of instantaneous heat recovery power fluctuation differences. Step S4-3: Analyze and process all the instantaneous heat recovery power fluctuation differences according to the three sigma principle, and select the calculated arithmetic mean as the center value. Subtract three times the standard deviation from the arithmetic mean to obtain the lower limit of fluctuation, and add three times the standard deviation to the arithmetic mean to obtain the upper limit of fluctuation. Based on the lower limit and the upper limit of fluctuation, the normal fluctuation range of instantaneous heat recovery power is formed. Step S4-4: Compare the instantaneous heat recovery power fluctuation difference corresponding to each operating condition feature portrait in the sliding window with the normal fluctuation range and the center value. The operating condition feature portraits whose instantaneous heat recovery power fluctuation difference equals the center value are assigned to the first operating condition feature portrait set; the operating condition feature portraits whose instantaneous heat recovery power fluctuation difference is greater than the lower fluctuation limit and less than the center value are assigned to the second operating condition feature portrait set; and the operating condition feature portraits whose instantaneous heat recovery power fluctuation difference is greater than the center value and less than the upper fluctuation limit are assigned to the third operating condition feature portrait set. The index of each operating condition feature portrait set is the instantaneous heat recovery power fluctuation difference, and the value is the operating condition feature portrait corresponding to the instantaneous heat recovery power fluctuation difference. In practice, the sliding window length of the operating condition feature profile is set based on the volume of historical data and the frequency of process fluctuations. Continuous historical operating condition feature profiles are extracted in chronological order according to timestamps. All instantaneous heat recovery power fluctuation differences within the window are extracted, and the arithmetic mean and standard deviation are calculated. The normal fluctuation range and center value, as well as the upper and lower limits of fluctuation, are established based on the three sigma principle. Each operating condition feature profile is classified into three categories of operating condition feature profile sets according to the comparison results of the power fluctuation difference and the interval boundary. The window sliding interval is kept consistent with the data acquisition cycle, and the interval judgment criteria are strictly followed to ensure the accuracy and rationality of the operating condition classification.

[0020] Step S5: Train the data of each working condition feature profile set through the neural network model to obtain the instantaneous heat recovery power fluctuation early warning model corresponding to the data division benchmark value. Collect real-time data of the process exhaust of the core process link of the chemical fiber plant, analyze and process it synchronously to obtain the real-time working condition feature profile, divide it based on the data division benchmark value, and provide early warning prompts for air conditioning heat recovery through the corresponding instantaneous heat recovery power fluctuation early warning model. Step S5-1: Take the first set of working condition feature profiles, the second set of working condition feature profiles, and the third set of working condition feature profiles as independent training datasets; perform feature vector transformation on the working condition feature profiles in each dataset, extract the temperature fluctuation feature data, humidity fluctuation feature data, air volume fluctuation feature data, and static pressure fluctuation feature data from each working condition feature profile, and combine them in a fixed order of temperature fluctuation feature, humidity fluctuation feature, air volume fluctuation feature, and static pressure fluctuation feature to construct a four-dimensional working condition feature vector; calculate the Euclidean modulus of each four-dimensional working condition feature vector, use the Euclidean modulus as the independent variable, and use the instantaneous heat recovery power fluctuation difference associated with the corresponding working condition feature profile as the dependent variable, and input them into three independent neural network models for training. After training, the first instantaneous heat recovery power fluctuation early warning model, the second instantaneous heat recovery power fluctuation early warning model, and the third instantaneous heat recovery power fluctuation early warning model are obtained. Step S5-2: Extract four feature data from the real-time working condition feature profile, combine them in the same fixed order to construct a real-time four-dimensional working condition feature vector, and calculate the real-time Euclidean modulus of the real-time four-dimensional working condition feature vector. Step S5-3: Compare the real-time instantaneous heat recovery power fluctuation difference corresponding to the real-time operating condition feature profile with the data division benchmark value, and match the corresponding instantaneous heat recovery power fluctuation early warning model. When the real-time instantaneous heat recovery power fluctuation difference is equal to the center value, the first early warning model is matched; when it is greater than the lower limit of the fluctuation but less than the center value, the second early warning model is matched; when it is greater than the center value but less than the upper limit of the fluctuation, the third early warning model is matched. Input the real-time Euclidean modulus into the matched instantaneous heat recovery power fluctuation early warning model, and output the instantaneous heat recovery power fluctuation prediction value corresponding to the first data acquisition timestamp in the next data acquisition cycle, which is recorded as the fluctuation prediction value. Compare the fluctuation prediction value with the normal fluctuation range of instantaneous heat recovery power. When the fluctuation prediction value is less than the lower limit of the fluctuation or greater than the upper limit of the fluctuation, an air conditioning heat recovery adjustment early warning prompt is issued; otherwise, no early warning prompt is triggered, and intelligent monitoring and management of air conditioning heat recovery continues. In practical implementation, the three types of operating condition feature profiles are used as independent training datasets. The four major fluctuation feature data are extracted from each profile and combined in a fixed order to construct a four-dimensional operating condition feature vector. The Euclidean modulus of the vector is calculated. The modulus is used as the independent variable and the corresponding power fluctuation difference is used as the dependent variable. The vectors are then input into independent neural network models to complete training. Real-time data from process exhaust and heat recovery devices are collected to construct real-time operating condition feature profiles and feature vectors. The real-time Euclidean modulus is calculated. The corresponding early warning model is matched according to the power fluctuation difference. The input modulus is used to obtain the fluctuation prediction value. The prediction value is compared with the normal fluctuation range to determine whether an early warning is triggered. The combination order of the feature vectors is kept consistent. The model is periodically iterated and optimized using new operating condition data to balance the sensitivity and accuracy of the early warning.

[0021] Example 2, as Figure 2 As shown, the present invention provides an intelligent management system for air conditioning heat recovery based on multi-source data fusion. The intelligent management system for air conditioning heat recovery includes an exhaust data acquisition module, a heat recovery feature integration module, an operating condition feature profile construction module, an operating condition set division module, and a heat recovery early warning and monitoring module. The exhaust data acquisition module is used to collect process exhaust characteristic parameters of the chemical fiber plant's manufacturing process and construct a process exhaust characteristic set; the heat recovery characteristic integration module is used to collect heat exchange parameters from the heat recovery device and combine them with normalized process exhaust characteristic parameters to construct a heat recovery characteristic set; the operating condition characteristic profile construction module is used to analyze historical heat recovery characteristic sets to obtain fluctuation differences and construct operating condition characteristic profiles based on the fluctuation difference set; the operating condition set division module is used to extract operating condition characteristic profiles through a sliding window, calculate a benchmark value based on the instantaneous heat recovery power fluctuation difference, and divide the operating condition characteristic profile set; the heat recovery early warning monitoring module is used to train an instantaneous heat recovery power fluctuation early warning model, predict based on real-time data, and trigger air conditioning heat recovery early warnings. The exhaust data acquisition module includes a parameter acquisition unit and a feature set construction unit. The parameter acquisition unit is used to continuously acquire the temperature, humidity, air volume and static pressure of the process exhaust air through deployed sensors according to a preset data acquisition cycle. The feature set construction unit is used to associate and bind each data acquisition timestamp with the corresponding process exhaust air feature parameter, and store them in a structured manner to form a process exhaust air feature set. The heat recovery feature integration module includes a heat exchange parameter acquisition unit and a feature set integration unit. The heat exchange parameter acquisition unit is used to simultaneously acquire the inlet temperature, outlet temperature, and flow rate of the recovery medium and calculate the instantaneous heat recovery power. The feature set integration unit is used to normalize the process exhaust characteristic parameters, then associate and bind them with the instantaneous heat recovery power, and store them in a structured manner to form a heat recovery feature set. The operating condition feature profile construction module includes a fluctuation difference calculation unit and a feature profile generation unit. The fluctuation difference calculation unit is used to calculate the instantaneous heat recovery power fluctuation difference and the difference of each sub-parameter of the process exhaust at adjacent data acquisition timestamps and associate them to form a heat recovery feature fluctuation difference set. The feature profile generation unit is used to parse the heat recovery feature fluctuation difference set, establish core feature dimensions and assign feature identifiers, and integrate them in a fixed order to form an operating condition feature profile. The operating condition set division module includes a sliding window interception unit and a feature set division unit; the sliding window interception unit is used to slide and intercept continuous historical operating condition feature profiles according to a preset window length and timestamp order; the feature set division unit is used to calculate the arithmetic mean and standard deviation of the instantaneous heat recovery power fluctuation difference, establish the normal fluctuation range, and divide the operating condition feature profiles into three sets. The heat recovery early warning and monitoring module includes an early warning model training unit and a real-time early warning and monitoring unit. The early warning model training unit uses three sets of operating condition feature profiles as training datasets, and trains three corresponding instantaneous heat recovery power fluctuation early warning models through feature vector transformation and Euclidean modulus calculation. The real-time early warning and monitoring unit collects real-time process exhaust and heat exchange parameters to construct real-time operating condition feature profiles, matches the predicted fluctuation values ​​of the corresponding early warning models, and determines whether to trigger an early warning after comparing them with the normal fluctuation range.

[0022] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of equivalents of the claims be included within the present invention.

Claims

1. An air conditioner heat recovery intelligent management method based on multi-source data fusion, characterized by: The intelligent management method for air conditioning heat recovery includes the following steps: Step S1: According to the preset data collection period, collect data on the process exhaust generated by the manufacturing process of the chemical fiber plant, obtain the process exhaust characteristic parameters, and construct a process exhaust characteristic set based on the data collection timestamps of each data collection period and the process exhaust characteristic parameters. Step S2: Synchronously collect heat exchange parameters from the heat recovery device side based on the data acquisition cycle to obtain instantaneous heat recovery power; read the process exhaust characteristic parameters in the corresponding process exhaust characteristic set according to the data acquisition timestamp as the associated index; after normalizing each process exhaust characteristic parameter, construct a heat recovery characteristic set in key-value pair storage form by combining the instantaneous heat recovery power. Step S3: Obtain the historical heat recovery feature set for data analysis to obtain the instantaneous heat recovery power fluctuation difference and the heat recovery feature fluctuation difference set between two adjacent data acquisition timestamps. Based on the heat recovery feature fluctuation difference set, analyze and extract features to construct the working condition feature profile corresponding to the instantaneous heat recovery power fluctuation difference. Step S4: Set up a sliding window for working condition feature profiles to extract and analyze data from multiple working condition feature profiles. Calculate the data division benchmark value based on the difference in instantaneous heat recovery power fluctuations extracted from multiple profiles. Divide the extracted working condition feature profiles into data based on the data division benchmark value to obtain the first working condition feature profile set, the second working condition feature profile set, and the third working condition feature profile set. Step S5: Train the neural network model on the feature profile set of each working condition to obtain the instantaneous heat recovery power fluctuation early warning model corresponding to the data division benchmark value. Collect real-time data on the process exhaust of the core process link of the chemical fiber plant, analyze and process it synchronously to obtain the real-time working condition feature profile, divide it based on the data division benchmark value, and provide early warning prompts for air conditioning heat recovery through the corresponding instantaneous heat recovery power fluctuation early warning model.

2. The intelligent management method for air conditioning heat recovery based on multi-source data fusion according to claim 1, characterized in that: The specific steps of step S1 are as follows: Step S1-1: By deploying temperature sensors, humidity sensors, air volume sensors, and static pressure sensors on the process exhaust ducts of the chemical fiber plant's manufacturing process, the characteristic parameters of the process exhaust are continuously collected according to a preset data acquisition cycle; the characteristic parameters of the process exhaust include the temperature, humidity, air volume, and static pressure of the process exhaust; the static pressure refers to the static pressure of the process exhaust within the process exhaust ducts of the chemical fiber plant's manufacturing process. Step S1-2: Associate and bind each data acquisition timestamp with the process exhaust characteristic parameters acquired at the corresponding time. Select the data acquisition timestamp as the key and the process exhaust characteristic parameter corresponding to the data acquisition timestamp as the value, and store them in a structured manner to form a process exhaust characteristic set.

3. The intelligent management method for air conditioning heat recovery based on multi-source data fusion according to claim 2, characterized in that: The specific steps of step S2 are as follows: Step S2-1: Based on a preset data acquisition cycle, synchronously acquire the heat exchange parameters on the heat recovery device side. The heat exchange parameters include the inlet temperature of the recovery medium, the outlet temperature of the recovery medium, and the flow rate of the recovery medium. The flow rate of the recovered medium is collected by deploying a flow meter sensor; the instantaneous heat recovery power is calculated based on the collected heat exchange parameters; Step S2-2: Select the data acquisition timestamp as the associated index, and read the process exhaust feature parameters corresponding to the data acquisition timestamp from the process exhaust feature set; retrieve the historical maximum and minimum values ​​corresponding to each sub-parameter in the process exhaust feature parameters, and perform normalization processing on each sub-parameter in the currently acquired process exhaust feature parameters. The normalization processing process is to subtract the corresponding historical minimum value from the current sub-parameter, and then divide by the difference between the corresponding historical maximum value and the historical minimum value. Step S2-3: Associate and bind the normalized process exhaust characteristic parameters with the instantaneous heat recovery power, select the data acquisition timestamp as the key, and select the parameter group composed of the normalized process exhaust characteristic parameters and the instantaneous heat recovery power as the value, and store it in a structured way to form a heat recovery feature set in key-value pair storage format.

4. The intelligent management method for air conditioning heat recovery based on multi-source data fusion according to claim 3, characterized in that: The specific steps of step S3 are as follows: Step S3-1: Sort the historical heat recovery feature set according to the order of data acquisition timestamps, extract the instantaneous heat recovery power corresponding to two adjacent data acquisition timestamps after sorting, calculate the difference between the instantaneous heat recovery power of the later data acquisition timestamp and the instantaneous heat recovery power of the earlier data acquisition timestamp, and record it as the instantaneous heat recovery power fluctuation difference. Step S3-2: Extract each sub-parameter from the normalized process exhaust characteristic parameters corresponding to two adjacent data acquisition timestamps after sorting, including temperature, humidity, air volume, and static pressure. Calculate the difference between the sub-parameters of the later data acquisition timestamp and the sub-parameters of the earlier data acquisition timestamp. Associate the temperature difference, humidity difference, air volume difference, and static pressure difference with the corresponding instantaneous heat recovery power fluctuation difference to form a heat recovery characteristic fluctuation difference set. The key of the heat recovery characteristic fluctuation difference set is the instantaneous heat recovery power fluctuation difference, and the value is a quadruple parameter group containing the temperature difference, humidity difference, air volume difference, and static pressure difference. Step S3-3: Perform structured analysis on each quaternary parameter group in the heat recovery characteristic fluctuation difference set to establish temperature difference, humidity difference, air volume difference, and static pressure difference as four independent core feature dimensions. A fixed feature identifier is assigned to each core feature dimension, where temperature difference corresponds to temperature fluctuation feature, humidity difference corresponds to humidity fluctuation feature, air volume difference corresponds to air volume fluctuation feature, and static pressure difference corresponds to static pressure fluctuation feature. Step S3-4: Map and associate the specific difference data corresponding to each core feature dimension with the feature identifier, and integrate them into a standardized data structure in a fixed order of temperature fluctuation feature, humidity fluctuation feature, air volume fluctuation feature, and static pressure fluctuation feature. This standardized data structure is the working condition feature profile corresponding to the instantaneous heat recovery power fluctuation difference.

5. The intelligent management method for air conditioning heat recovery based on multi-source data fusion according to claim 4, characterized in that: The specific steps of step S4 are as follows: Step S4-1: Based on the window length of the preset working condition feature profile sliding window, and using the number of data acquisition timestamps as the unit of measurement, slide and cut the continuous historical working condition feature profiles according to the order of the data acquisition timestamps in the historical heat recovery feature set. After each sliding cut, the window moves backward at the interval of a single data acquisition timestamp until the cutting operation of all working condition feature profiles is completed. Step S4-2: Extract all instantaneous heat recovery power fluctuation differences contained in the sliding window of each captured working condition feature image, calculate the corresponding arithmetic mean of the extracted instantaneous heat recovery power fluctuation differences, and calculate the standard deviation of the set of instantaneous heat recovery power fluctuation differences.

6. The intelligent management method for air conditioning heat recovery based on multi-source data fusion according to claim 5, characterized in that: Step S4 also includes: Step S4-3: Analyze and process all the instantaneous heat recovery power fluctuation differences according to the three sigma principle, and select the calculated arithmetic mean as the center value. Subtract three times the standard deviation from the arithmetic mean to obtain the lower limit of fluctuation, and add three times the standard deviation to the arithmetic mean to obtain the upper limit of fluctuation. Based on the lower limit and the upper limit of fluctuation, the normal fluctuation range of instantaneous heat recovery power is formed. Step S4-4: Compare the instantaneous heat recovery power fluctuation difference corresponding to each operating condition feature portrait in the sliding window with the normal fluctuation range and the center value. The operating condition feature portrait with an instantaneous heat recovery power fluctuation difference equal to the center value is assigned to the first operating condition feature portrait set; the operating condition feature portrait with an instantaneous heat recovery power fluctuation difference greater than the lower fluctuation limit and less than the center value is assigned to the second operating condition feature portrait set; and the operating condition feature portrait with an instantaneous heat recovery power fluctuation difference greater than the center value and less than the upper fluctuation limit is assigned to the third operating condition feature portrait set. The index of each operating condition feature portrait set is the instantaneous heat recovery power fluctuation difference, and the value is the operating condition feature portrait corresponding to the instantaneous heat recovery power fluctuation difference.

7. The intelligent management method for air conditioning heat recovery based on multi-source data fusion according to claim 6, characterized in that: The specific steps of step S5 are as follows: Step S5-1: Take the first set of working condition feature profiles, the second set of working condition feature profiles, and the third set of working condition feature profiles as independent training datasets; perform feature vector transformation on the working condition feature profiles in each dataset, extract the temperature fluctuation feature data, humidity fluctuation feature data, air volume fluctuation feature data, and static pressure fluctuation feature data from each working condition feature profile, and combine them in a fixed order of temperature fluctuation feature, humidity fluctuation feature, air volume fluctuation feature, and static pressure fluctuation feature to construct a four-dimensional working condition feature vector; calculate the Euclidean modulus of each four-dimensional working condition feature vector, use the Euclidean modulus as the independent variable, and use the instantaneous heat recovery power fluctuation difference associated with the corresponding working condition feature profile as the dependent variable, and input them into three independent neural network models for training. After training, the first instantaneous heat recovery power fluctuation early warning model, the second instantaneous heat recovery power fluctuation early warning model, and the third instantaneous heat recovery power fluctuation early warning model are obtained. Step S5-2: Extract four feature data from the real-time working condition feature profile, combine them in the same fixed order to construct a real-time four-dimensional working condition feature vector, and calculate the real-time Euclidean modulus of the real-time four-dimensional working condition feature vector. Step S5-3: Compare the real-time instantaneous heat recovery power fluctuation difference corresponding to the real-time operating condition feature profile with the data division benchmark value, and match the corresponding instantaneous heat recovery power fluctuation early warning model. When the real-time instantaneous heat recovery power fluctuation difference is equal to the center value, the first early warning model is matched; when it is greater than the lower limit of the fluctuation but less than the center value, the second early warning model is matched; when it is greater than the center value but less than the upper limit of the fluctuation, the third early warning model is matched. Input the real-time Euclidean modulus into the matched instantaneous heat recovery power fluctuation early warning model, and output the instantaneous heat recovery power fluctuation prediction value corresponding to the first data acquisition timestamp in the next data acquisition cycle, which is recorded as the fluctuation prediction value. The predicted fluctuation value is compared with the normal fluctuation range of instantaneous heat recovery power. When the predicted fluctuation value is less than the lower limit of fluctuation or greater than the upper limit of fluctuation, an early warning prompt for air conditioning heat recovery adjustment is issued. Conversely, if no warning is triggered, intelligent monitoring and management of air conditioning heat recovery will continue.

8. An intelligent management system for air conditioning heat recovery based on multi-source data fusion, applied to the intelligent management method for air conditioning heat recovery based on multi-source data fusion as described in any one of claims 1-7, characterized in that: The intelligent management system for air conditioning heat recovery includes an exhaust data acquisition module, a heat recovery feature integration module, an operating condition feature profile construction module, an operating condition set division module, and a heat recovery early warning and monitoring module. The exhaust data acquisition module is used to collect process exhaust characteristic parameters of the chemical fiber plant's manufacturing process and construct a process exhaust characteristic set; the heat recovery characteristic integration module is used to collect heat exchange parameters from the heat recovery device and combine them with normalized process exhaust characteristic parameters to construct a heat recovery characteristic set; the operating condition characteristic profile construction module is used to analyze historical heat recovery characteristic sets to obtain fluctuation differences, and construct an operating condition characteristic profile based on the fluctuation difference set; the operating condition set division module is used to extract the operating condition characteristic profile through a sliding window, calculate the benchmark value based on the instantaneous heat recovery power fluctuation difference, and divide the operating condition characteristic profile set; the heat recovery early warning monitoring module is used to train an instantaneous heat recovery power fluctuation early warning model, predict based on real-time data, and trigger air conditioning heat recovery early warning.

9. The intelligent management system for air conditioning heat recovery based on multi-source data fusion according to claim 8, characterized in that: The exhaust data acquisition module includes a parameter acquisition unit and a feature set construction unit; the parameter acquisition unit is used to continuously acquire the temperature, humidity, air volume and static pressure of the process exhaust air according to a preset data acquisition cycle through deployed sensors; The feature set construction unit is used to associate and bind each data acquisition timestamp with the corresponding process exhaust feature parameter, and store them in a structured manner to form a process exhaust feature set; The heat recovery feature integration module includes a heat exchange parameter acquisition unit and a feature set integration unit. The heat exchange parameter acquisition unit is used to simultaneously acquire the inlet temperature, outlet temperature, and flow rate of the recovery medium and calculate the instantaneous heat recovery power. The feature set integration unit is used to normalize the process exhaust characteristic parameters, then associate and bind them with the instantaneous heat recovery power, and store them in a structured manner to form a heat recovery feature set. The working condition feature profile construction module includes a fluctuation difference calculation unit and a feature profile generation unit; The fluctuation difference calculation unit is used to calculate the instantaneous heat recovery power fluctuation difference between adjacent data acquisition timestamps and the difference between each sub-parameter of the process exhaust, and associate them to form a set of heat recovery characteristic fluctuation differences; The feature profile generation unit is used to analyze the set of heat recovery feature fluctuation differences, establish core feature dimensions and assign feature identifiers, and integrate them in a fixed order to form a working condition feature profile.

10. The intelligent management system for air conditioning heat recovery based on multi-source data fusion according to claim 8, characterized in that: The working condition set division module includes a sliding window truncation unit and a feature set division unit; the sliding window truncation unit is used to slide and truncate continuous historical working condition feature profiles according to a preset window length and timestamp order; The feature set partitioning unit is used to calculate the arithmetic mean and standard deviation of the instantaneous heat recovery power fluctuation difference, establish the normal fluctuation range, and divide the working condition feature profile into three sets. The heat recovery early warning and monitoring module includes an early warning model training unit and a real-time early warning and monitoring unit. The early warning model training unit uses three sets of operating condition feature profiles as training datasets, and trains three corresponding instantaneous heat recovery power fluctuation early warning models through feature vector transformation and Euclidean modulus calculation. The real-time early warning and monitoring unit collects real-time process exhaust and heat exchange parameters to construct real-time operating condition feature profiles, matches the predicted fluctuation values ​​of the corresponding early warning models, and determines whether to trigger an early warning after comparing them with the normal fluctuation range.

Images