Customized energy-saving air conditioner control method and device oriented to industrial process requirements

Abstract

The invention provides a customized energy-saving air conditioner control method and device for industrial process requirements, and is applied to the technical field of data processing. According to the method, electromagnetic interference and vibration noise are eliminated through Kalman filtering, and a standardized process-environment-energy consumption correlation sequence is generated; and converting the three-dimensional process load map into a three-dimensional process load map, and establishing an individualized air conditioner dynamic load prediction model in combination with heat and humidity characteristics of a scene through fluid dynamic simulation and self-adaptive grid division. Extracting process priority factors to construct an adjustment matrix, calculating an air conditioner parameter combination under target energy consumption, extracting energy consumption characteristics, dividing standard exceeding risk levels, and constructing a multi-dimensional characteristic matrix; and comparing real-time data with historical data to identify abnormity, and generating an energy consumption optimization correction factor. Users are grouped according to enterprise conditions, key factors are screened by using a gradient boosting tree, a personalized energy-saving control model is constructed by fusing multiple information, target parameters are output, and a real-time adjustment instruction and a time-phased energy-saving strategy are generated in combination with a time sequence.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a customized energy-saving air conditioning control method and device for industrial process needs. Background Technology

[0002] In industrial production, air conditioning systems are crucial equipment for ensuring process stability, and their energy consumption accounts for a significant proportion of total industrial energy consumption. Therefore, achieving energy-saving control while meeting process requirements is of great importance. However, existing industrial air conditioning control technologies have the following shortcomings:

[0003] Industrial environments contain a large amount of electromagnetic interference (such as electromagnetic signals generated by machine tools, frequency converters and other equipment) and equipment vibration noise (such as the vibration of punch presses and fans during operation). These interferences can cause the collected process parameters, environmental data and energy consumption data to be distorted, making the basic data on which air conditioning control is based inaccurate, which in turn affects the adjustment accuracy and makes it difficult to meet the stringent temperature and humidity requirements of high-precision processes (such as electronic component welding, food baking and other scenarios).

[0004] Existing air conditioning load prediction models are mostly designed based on general scenarios, without fully considering the differences in thermal and humidity characteristics of different industrial processes (such as the drastically different thermal and humidity requirements of machining workshops and textile printing and dyeing workshops). Furthermore, there is insufficient detailed analysis of high-load process areas, resulting in a large deviation between the prediction results and the actual load, which fails to provide accurate basis for energy-saving control.

[0005] The lack of a systematic assessment of the risk of excessive energy consumption makes it difficult to identify abnormal situations such as short-term energy consumption surges and load-energy consumption imbalances in advance. At the same time, the absence of a dynamic correction mechanism for energy consumption optimization based on real-time operating conditions means that when process parameters or environmental conditions change, air conditioning operating parameters cannot be adjusted in a timely manner, resulting in energy waste or failure to meet process requirements.

[0006] Industrial enterprises vary significantly in terms of production scale, process complexity, and equipment configuration (e.g., the air conditioning needs of small-scale machining enterprises differ from those of large-scale automobile manufacturing enterprises). Existing control models mostly use uniform parameter settings and are not customized for the characteristics of different user groups, making it difficult to balance energy-saving effects and process adaptability.

[0007] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0008] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part by practice of the invention.

[0009] According to one aspect of this application, a customized energy-saving air conditioning control method for industrial process needs is provided, comprising: acquiring real-time process parameters, environmental data, and historical energy consumption records of air conditioning operation in industrial production; using a Kalman filter algorithm to remove electromagnetic interference and equipment vibration noise in the industrial environment, generating a standardized process-environment-energy consumption correlation data sequence; converting the standardized process-environment-energy consumption correlation data sequence into a three-dimensional process load distribution map, processing it with fluid dynamics simulation software, subdividing high-load process areas using an adaptive mesh partitioning algorithm, and constructing an individualized air conditioning dynamic load prediction model based on the differences in thermal and humidity characteristics of different industrial scenarios; extracting priority factors for different industrial processes based on the individualized air conditioning dynamic load prediction model to construct a dynamic parameter adjustment matrix, performing parallel calculations on the operating status of multiple air conditioning modules, and obtaining the air conditioning operating parameter combination under the target energy consumption; extracting energy consumption peak values ​​according to industrial process scenarios, Load fluctuation rate and process-energy consumption matching difference value are used to classify the risk level of energy consumption exceeding the standard, and a multi-dimensional feature matrix containing process demand intensity, environmental interference coefficient, and equipment energy efficiency degradation information is constructed. Real-time energy consumption is compared with historical data of the same process to identify abnormal signals such as short-term energy consumption surges and load-energy consumption imbalance. The slope of the energy consumption-process parameter correlation curve is used to generate a dynamic correction factor for energy consumption optimization. Users are grouped according to the production scale, process complexity, and equipment configuration of industrial enterprises. The gradient boosting tree algorithm is used to screen key influencing factors of the multi-dimensional feature matrix and the dynamic correction factor for energy consumption optimization. The energy consumption optimization parameters, process constraints, and equipment operating limit information are integrated to construct a personalized energy-saving control model. Based on the target operating parameter output of the personalized energy-saving control model, combined with the time correlation information of industrial process timing and air conditioning equipment operating characteristics, real-time dynamic adjustment instructions and time-sharing energy-saving control strategies are generated.

[0010] Another aspect of this application discloses a customized energy-saving air conditioning control device for industrial process needs, comprising: an acquisition module for acquiring real-time process parameters, environmental data, and historical energy consumption records of air conditioning operation in industrial production; employing a Kalman filter algorithm to remove electromagnetic interference and equipment vibration noise from the industrial environment; and generating a standardized process-environment-energy consumption correlation data sequence; a processing module for converting the standardized process-environment-energy consumption correlation data sequence into a three-dimensional process load distribution map; processing the map using fluid dynamics simulation software; subdividing high-load process areas using an adaptive mesh partitioning algorithm; and constructing an individualized air conditioning dynamic load prediction model based on the differences in thermal and humidity characteristics of different industrial scenarios; extracting priority factors for different industrial processes based on the individualized air conditioning dynamic load prediction model to construct a dynamic parameter adjustment matrix; performing parallel calculations on the operating states of multiple air conditioning modules to obtain the air conditioning operating parameter combination under the target energy consumption; and extracting parameters according to industrial process scenarios. Energy consumption peak, load fluctuation rate, and process-energy consumption matching difference value are used to classify the risk level of energy consumption exceeding the standard, and a multi-dimensional feature matrix containing process demand intensity, environmental interference coefficient, and equipment energy efficiency degradation information is constructed. Real-time energy consumption is compared with historical data of the same process to identify abnormal signals such as short-term energy consumption surges and load-energy consumption imbalance. The slope of the energy consumption-process parameter correlation curve is used to generate a dynamic correction factor for energy consumption optimization. Users are grouped according to the production scale, process complexity, and equipment configuration of industrial enterprises. The gradient boosting tree algorithm is used to screen key influencing factors of the multi-dimensional feature matrix and the dynamic correction factor for energy consumption optimization. A personalized energy-saving control model is constructed by integrating energy consumption optimization parameters, process constraints, and equipment operating limit information. Based on the target operating parameter output of the personalized energy-saving control model, combined with the time correlation information of industrial process timing and air conditioning equipment operating characteristics, real-time dynamic adjustment instructions and time-sharing energy-saving control strategies are generated.

[0011] According to another aspect of this application, an electronic device includes: a first processor; and a memory for storing executable instructions of the first processor; wherein the first processor is configured to execute the above-described customized energy-saving air conditioning control method for industrial process requirements by executing the executable instructions.

[0012] According to another aspect of this application, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a second processor, implements the above-described customized energy-saving air conditioning control method for industrial process requirements.

[0013] This application provides a customized energy-saving air conditioning control method and device for industrial process needs. It processes data using Kalman filtering to generate standardized correlation sequences and convert them into a three-dimensional load map. An individualized load prediction model is constructed through fluid dynamics simulation and adaptive mesh generation. Process priority factors are extracted, and parameter combinations under target energy consumption are obtained through parallel computing. Energy consumption risk levels are classified, and dynamic correction factors are generated. Users are grouped according to their enterprise situation, and key factors are selected to construct personalized control models. Finally, real-time adjustment commands and time-segmented strategies are generated. This effectively eliminates industrial environmental interference and improves data accuracy; achieves accurate load prediction and adapts to the thermal and humidity characteristics of different scenarios; reduces energy waste through risk assessment and dynamic correction; the personalized model balances energy saving and process requirements, and the time-segmented strategy further optimizes energy consumption, significantly improving the energy efficiency of the air conditioning system while ensuring the stability of industrial production.

[0014] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0015] Figure 1 A flowchart illustrating a customized energy-saving air conditioning control method for industrial process needs provided in an embodiment of this application is shown.

[0016] Figure 2 This illustration shows a structural schematic diagram of a customized energy-saving air conditioning control device for industrial process needs, provided in one embodiment of this application. Detailed Implementation

[0017] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0018] The following combination Figure 1 This application describes a customized energy-saving air conditioning control method for industrial process needs, based on exemplary embodiments thereof. It should be noted that the application scenarios described below are merely illustrative for understanding the spirit and principles of this application, and the embodiments of this application are not limited in any way. Rather, the embodiments of this application are applicable to any suitable scenario.

[0019] In one embodiment, this application also proposes a customized energy-saving air conditioning control method and device for industrial process needs. Figure 1 A schematic flowchart of a customized energy-saving air conditioning control method for industrial process needs, according to an embodiment of this application, is shown.

[0020] S101 acquires real-time process parameters, environmental data, and historical energy consumption records of air conditioning operation in industrial production. It uses a Kalman filter algorithm to remove electromagnetic interference and equipment vibration noise in the industrial environment, and generates a standardized process-environment-energy consumption correlation data sequence.

[0021] In one implementation, real-time process parameters in industrial production are indicators reflecting the core state of the industrial production process. Key parameters need to be determined and collected in real-time according to different industrial scenarios. Taking machining as an example, the real-time process parameters to be collected include machine tool cutting speed (m / min), feed rate (mm / r), and depth of cut (mm). These parameters directly affect the heat generation in the machining area, thus affecting air conditioning load demand. In electronic component welding scenarios, real-time process parameters include welding temperature (°C), welding time (s), and solder supply (g / s). Excessively high welding temperatures can cause a sudden rise in local ambient temperature, requiring timely adjustment by the air conditioning system. Data collection can be achieved through real-time monitoring using industrial sensors, such as installing speed sensors on the machine tool spindle and temperature sensors at the welding station. Sensor data is transmitted in real-time to the data acquisition terminal via an industrial bus (Profinet).

[0022] Environmental data collection needs to focus on key environmental factors affecting air conditioning operation within industrial spaces to ensure the data reflects the actual state of the industrial environment. Taking an automotive painting workshop as an example, the environmental data to be collected includes workshop temperature (°C), relative humidity (%RH), and air cleanliness (particles / m³, counting particles ≥0.5μm). Temperature and humidity directly affect the drying effect of the paint, while air cleanliness relates to the coating quality. These data all need to be adjusted by the air conditioning system according to actual conditions. In food processing workshops, environmental data also needs to include workshop air pressure (Pa) to prevent external pollutants from entering the clean processing area. During data collection, temperature and humidity sensors, particle counters, and air pressure sensors can be deployed in different areas of the workshop (such as painting stations, drying areas, and workshop corners). The sensors collect data every 10 seconds to ensure real-time reflection of environmental changes.

[0023] Historical energy consumption records for air conditioning systems need to collect energy consumption data from past operations to provide historical reference for subsequent energy consumption analysis and optimization. Taking a central air conditioning energy-saving project in a machinery manufacturing company as an example, the required historical energy consumption records include daily power consumption (kWh) of the air conditioning system over the past 12 months, energy consumption data under different operating modes (such as cooling, heating, and ventilation), and individual energy consumption records for each air conditioning module (such as compressor, fan, and water pump). Simultaneously, it needs to be linked to production process information (such as daily product types and production duration) and environmental data (such as daily average temperature and humidity) for the corresponding time period to clarify the historical correlation between energy consumption and processes and the environment. The data is obtained by exporting from the historical database of the air conditioning control system. If the air conditioning system does not have built-in historical data storage capabilities, smart meters and energy monitoring modules can be installed to retrospectively collect energy consumption data over a period of time, ensuring that the data span is long enough to cover energy consumption under different seasons and production loads.

[0024] Industrial environments are rife with electromagnetic interference (such as electromagnetic signals generated by machine tools and frequency converters) and equipment vibration noise (such as sensor data fluctuations caused by the vibration of punch presses and fans). This interference can distort the collected data, necessitating Kalman filtering to process the data, eliminate interference, and ensure data accuracy. Taking the air conditioning data processing in a steel rolling mill workshop as an example, the operation of the rolling mill generates strong electromagnetic interference, causing irregular fluctuations in temperature data collected by temperature and humidity sensors (e.g., when the actual temperature is stable at 28℃, the sensor data fluctuates randomly between 26-30℃). Simultaneously, the vibration of fans and rolling mills in the workshop can cause instantaneous abnormal values ​​in the power consumption data collected by energy consumption sensors (e.g., when the normal energy consumption is 50kWh / h, occasional instantaneous data of 80kWh / h appears).

[0025] When using the Kalman filter algorithm, the system state equation and observation equation are first established: the actual values ​​of temperature and energy consumption are taken as the system state, and the interference-laden data collected by the sensors are taken as the observation values. Then, through a prediction step, the system state and error covariance at the current moment are predicted based on the optimal estimate from the previous moment. Next, through an update step, the Kalman gain is calculated by combining the current observation value and the prediction error to correct the predicted value, thus obtaining the optimal estimate for the current moment. For example, for temperature data, after Kalman filtering, the original fluctuation of 26-30℃ is corrected to an accurate value stable at around 28℃, with the fluctuation range controlled within ±0.5℃. For energy consumption data, the instantaneous abnormal 80kWh / h data is corrected to a reasonable range of 50-52kWh / h, effectively eliminating the influence of electromagnetic interference and vibration noise.

[0026] After data acquisition and interference removal, process parameters, environmental data, and energy consumption data need to be standardized, and the correlation between them needs to be established to form a standardized process-environment-energy consumption correlation data sequence. This ensures that the data format is uniform and the logical relationship is clear, facilitating subsequent graph conversion and model construction. Data standardization requires converting data of different dimensions and units into standard data of a uniform magnitude to eliminate the influence of dimensions. Taking the air conditioning data of an electronics factory as an example, in the original data, the process parameter welding temperature is 220-250℃, the environmental data workshop temperature is 25-30℃, and the energy consumption data air conditioning power consumption is 80-120kWh / h. The units and numerical ranges of the three vary greatly. The Min-Max standardization method is used to map all data to the [0,1] interval. The standardization formula is: Standardized data = (Original data - Minimum data value) / (Maximum data value - Minimum data value). For example, a welding temperature of 220℃ corresponds to a standardized value of 0, 250℃ corresponds to a standardized value of 1, and 235℃ is calculated as (235-220) / (250-220)=0.5; a workshop temperature of 25℃ corresponds to 0, 30℃ corresponds to 1, and 27.5℃ is 0.5 after standardization; an air conditioning power consumption of 80kWh / h corresponds to 0, 120kWh / h corresponds to 1, and 100kWh / h is 0.5 after standardization.

[0027] Establish the correlation between process, environmental, and energy consumption data based on the time dimension, ensuring that the three types of data at the same time point correspond one-to-one, forming a correlated data sequence. Taking the collected and processed data at a time interval of 1 hour as an example, in the standardized data of a certain moment (e.g., 9:00-10:00 on May 10, 2024), the standardized value of the process parameter welding temperature is 0.5, the standardized value of the environmental data workshop temperature is 0.5, and the standardized value of the energy consumption data air conditioning power consumption is 0.5. These three data are correlated in the order of "process parameter - environmental data - energy consumption data" to form the correlated data group (0.5, 0.5, 0.5) for that moment. The correlation is then formed by sequentially organizing data from different moments (e.g., 9:00-10:00 on May 10, 2024). The associated data groups (e.g., 00-10:00, 10:00-11:00, 11:00-12:00, etc.) are arranged in chronological order to generate a standardized process-environment-energy consumption correlation data sequence, such as [(0.5,0.5,0.5),(0.6,0.6,0.6),(0.4,0.4,0.4),...]. This sequence can clearly reflect the correspondence between process, environment, and energy consumption at different time points, laying the foundation for subsequent conversion into a three-dimensional process load distribution map.

[0028] S102, after converting the standardized process-environment-energy consumption correlation data sequence into a three-dimensional process load distribution map, it is processed by fluid dynamics simulation software, and an adaptive grid partitioning algorithm is selected to subdivide the high-load process area. Combined with the differences in thermal and humidity characteristics of different industrial scenarios, an individualized air conditioning dynamic load prediction model is constructed.

[0029] In one implementation, a standardized process-environment-energy consumption correlation data sequence is converted into a three-dimensional process load distribution map to generate basic load visualization data. When constructing the three-dimensional process load distribution map, the data in the standardized process-environment-energy consumption correlation data sequence must first be processed according to a defined dimensional mapping rule: process parameters are mapped to the X-axis, environmental data to the Y-axis, and energy consumption data to the Z-axis, thereby constructing a multi-dimensional correlation visualization map and generating basic load visualization data. Under this rule, each data group in the sequence will correspond to a data point in the map. Taking an electronic component welding workshop as an example, in its standardized process-environment-energy consumption correlation data sequence, a data group for a certain period is (standardized welding temperature value 0.6, standardized workshop temperature value 0.5, standardized air conditioning power consumption value 0.7). According to the above mapping rule, the coordinates of this data group in the three-dimensional map are (X=0.6, Y=0.5, Z=0.7). By mapping the coordinates of all data sets in the sequence (such as (0.5, 0.4, 0.6), (0.7, 0.6, 0.8), etc.), a complete three-dimensional process load distribution map is formed. In this map, a higher Z-axis value represents higher energy consumption, while changes in the X-axis and Y-axis values ​​reflect fluctuations in process parameters and environmental conditions, respectively. By observing the map, a correlation can be intuitively observed: when welding temperature increases (X-axis increases) + workshop temperature increases (Y-axis increases), energy consumption (Z-axis) usually increases accordingly. This visualization method makes the basic load data more intuitive and helps to quickly locate the specific process and environmental combinations that lead to high energy consumption.

[0030] The three-dimensional process load distribution map is imported into fluid dynamics simulation software for simulation calculations to generate thermal and moisture load field simulation data. To generate this data, the three-dimensional process load distribution map needs to be imported into fluid dynamics simulation software (such as Fluent or STAR-CCM+) for simulation calculations. During the simulation, the fluid dynamics simulation software simulates the airflow trajectory and heat and moisture exchange efficiency within the industrial space, and combines this with the load distribution in the three-dimensional map to quantify the thermal and moisture load intensity of different process areas. Finally, it calculates and outputs the thermal and moisture load values ​​for each process area. Taking an automotive painting workshop as an example, firstly, the three-dimensional process load distribution map of the workshop (X-axis for painting line operating speed, Y-axis for workshop temperature and humidity, and Z-axis for air conditioning energy consumption) is imported into Fluent software. Next, a physical model identical to the actual workshop (40m × 20m × 8m in size, including 3 painting lines and 2 drying zones) is built in the software. Then, the data from the three-dimensional map is precisely correlated to the corresponding areas of the physical model (e.g., the area where painting line 1 has an operating speed of X=0.8 and workshop humidity of Y=0.6 corresponds to high energy consumption data of Z=0.9). After starting the software simulation calculation, the software simulates the complete airflow trajectory from the air inlet to the air outlet in the workshop (e.g., vortices are formed near the drying zone, resulting in a longer air residence time), while quantifying the heat and humidity exchange efficiency of different areas (e.g., the heat exchange efficiency of the drying zone reaches 90%, while the heat exchange efficiency of the ordinary painting zone is about 60%). Finally, simulated data of the heat and humidity load field were generated. The data clearly showed that the heat and humidity load intensity of the drying zone was 80W / ㎡, which was much higher than the 40W / ㎡ of the ordinary coating zone, clearly showing the load difference between different process areas.

[0031] An adaptive mesh generation algorithm is used to subdivide the high-load process areas in the thermal and humidity load field simulation data, generating differentiated mesh data. To accurately analyze the thermal and humidity load field simulation data and distinguish load differences between different areas, an adaptive mesh generation algorithm is used to subdivide the high-load process areas in the simulation data, thereby generating differentiated mesh data. During this process, the adaptive mesh generation algorithm identifies the thermal and humidity load density in the simulation data in real time. When the load density exceeds a preset threshold, a mesh refinement mechanism is automatically triggered. Simultaneously, mesh resources are configured according to the core rule of "using high-density meshes for high-load areas and low-density meshes for low-load areas." In specific implementation, the load density threshold must first be set through the algorithm, and then the mesh generation operation is carried out based on this threshold. Taking a machining workshop as an example, the thermal and humidity load density threshold is first set to 60W / ㎡. Then, the algorithm identifies the load density in the thermal and humidity load field simulation data in real time: from the data, it is clear that the lathe machining area, as a high-load process area, has a load density of 75W / ㎡ (exceeding the preset threshold), while the warehouse area has a load density of 30W / ㎡ (below the preset threshold). Based on the aforementioned grid division rules, the algorithm triggers grid refinement for the lathe machining area, configuring it with a high-density grid of 2cm×2cm×2cm. This precision grid can accurately capture local load fluctuations in the lathe machining area caused by machine tool cutting heat (e.g., load density reaches 85W / ㎡ near the spindle, and 65W / ㎡ at the edge). For the low-load warehouse area, a low-density grid of 10cm×10cm×10cm is used, which only needs to reflect the overall stable low-load state of the warehouse area. In the final generated differentiated grid data, the high-load lathe machining area has 5000 grids, while the low-load warehouse area has only 800 grids. While ensuring the calculation accuracy of the high-load area, the overall data volume is significantly reduced, providing efficient and accurate data support for the subsequent construction of a personalized air conditioning dynamic load prediction model.

[0032] To accurately adapt to the air conditioning control needs of different industrial scenarios, it is necessary to first collect the thermal and humidity characteristic difference parameters of different industrial scenarios, and then build a scenario-based thermal and humidity characteristic constraint library based on these parameters, thereby generating scenario constraint data. Taking food baking workshops and textile printing and dyeing workshops as examples: For food baking workshops, the thermal and humidity characteristics of the dough fermentation stage are collected first—the temperature needs to be stable at 28-30℃, the relative humidity needs to be maintained at 70%-75%, and excessive temperature fluctuations will lead to uneven dough fermentation (maximum tolerable fluctuation ±1℃), while excessive humidity fluctuations will affect the quality of the dough (maximum tolerable fluctuation ±3%RH). These parameters are recorded as specific constraints for this scenario in the scenario-based thermal and humidity characteristic constraint library. For textile printing and dyeing workshops, the thermal and humidity characteristics of the fabric dyeing stage are collected—the temperature needs to be controlled at 60-65℃, and the relative humidity needs to be maintained at 50%-55%. Because the fabric has strong high temperature resistance (temperature fluctuation range ±2℃), the impact of humidity fluctuations on dyeing uniformity is relatively mild (allowable range ±5%RH). At the same time, it was found that "a sudden drop in humidity exceeding 5%RH / h during the dyeing process will lead to uneven dyeing of the fabric," so this special constraint is added and also recorded in the constraint library. In the final constructed scenario-based thermal and humidity characteristic constraint library, each industrial scenario corresponds to a set of exclusive constraint parameters. The generated scenario constraint data can clearly define the thermal and humidity boundary conditions that the air conditioning system needs to meet under different scenarios, providing a scenario-based constraint basis for the subsequent construction of individualized air conditioning dynamic load prediction models.

[0033] By integrating differentiated grid data and scenario-constrained data, an individualized dynamic load prediction model for air conditioning is constructed. The model construction requires first extracting load characteristics (such as load intensity and fluctuation patterns of each grid) from the differentiated grid data, and then incorporating the boundary conditions from the scenario-constrained data to establish a "load-scenario constraint" correlation prediction logic. Taking a food baking workshop as an example: First, the load characteristics of the high-density grid in the fermentation zone are extracted from the differentiated grid data—the load intensity of the fermentation zone is stable at 55-60 W / m² during the day from 8:00 to 18:00 (due to continuous heat generation from fermentation), and the load drops to 30-35 W / m² from 18:00 to 8:00 the next day (fermentation is suspended); then, the conditions of "temperature 28-30℃, humidity 70%-75%, temperature fluctuation ±1℃" in the scenario constraint data are incorporated to establish the prediction logic: when the model predicts that the load in the fermentation zone will rise to 58 W / m², if the current workshop temperature has reached 29.5℃ (close to the upper limit of the constraint), then the air conditioning load needs to be increased by 10% to control the temperature not to exceed 30℃; if the current temperature is 28.5℃, then the air conditioning load needs to be increased by 5% to maintain temperature stability. Through this integration, the individualized dynamic load prediction model for air conditioning can accurately predict the air conditioning load demand in different time periods and areas based on the grid load changes and thermal and humidity constraints of the baking workshop. For example, it can predict that the air conditioning load needs to be maintained at 120kW during the peak fermentation period during the day, while it can be reduced to 80kW at night.

[0034] S103 extracts priority factors of different industrial processes based on the individualized air conditioning dynamic load prediction model to construct a dynamic parameter adjustment matrix, performs parallel calculations on the operating status of multiple air conditioning modules, and obtains the combination of air conditioning operating parameters under the target energy consumption.

[0035] In one implementation, priority factors for different industrial processes are extracted based on a personalized air conditioning dynamic load prediction model, and a dynamic parameter adjustment matrix is ​​constructed. The priority factor extraction uses the production criticality, urgency of heat and humidity demand, and energy consumption sensitivity of the industrial processes as core dimensions. The dynamic parameter adjustment matrix maps each process priority factor to air conditioning adjustment parameters, generating a parameter adjustment logic framework that can be updated in real time. To achieve precise dynamic adjustment of air conditioning parameters, priority factors for different industrial processes need to be extracted based on the personalized air conditioning dynamic load prediction model, and a dynamic parameter adjustment matrix needs to be constructed. Specifically, the priority factor extraction uses the production criticality, urgency of heat and humidity demand, and energy consumption sensitivity of the industrial processes as core dimensions, and needs to combine the load characteristics output by the personalized air conditioning dynamic load prediction model (such as the load intensity and fluctuation patterns of each process area) to quantify and assign values ​​to each core dimension. Taking an automotive parts production workshop as an example, this workshop includes three types of processes: engine block processing (process A), parts cleaning (process B), and finished product assembly (process C). Based on the load characteristics output by the model, the specific quantification and assignment of each dimension are as follows:

[0036] Production Criticality: The load characteristics of process A (engine block machining) show that it is extremely sensitive to temperature fluctuations. Temperature deviations will directly lead to the failure of machining accuracy, which in turn will cause the entire engine to be scrapped. The production criticality value is assigned to 0.9 (out of 1.0); The load characteristics of process B (cleaning) show that its humidity fluctuations only indirectly affect the cleanliness of parts, and have a moderate impact on the quality of the final product. The production criticality value is assigned to 0.6; The load characteristics of process C (assembly) show that its requirements for temperature and humidity are relaxed. The production criticality value is the lowest, assigned to 0.3.

[0037] Urgency of heat and humidity requirements: The load characteristics of process A indicate that it needs to maintain a workshop temperature of 20±1℃ and a relative humidity of 50±5%. A temperature deviation of 1℃ will trigger a load anomaly, requiring immediate adjustment. The urgency of heat and humidity requirements is assigned a value of 0.8. The load characteristics of process B allow for short-term (e.g., within 10 minutes) small fluctuations in temperature within the range of 25±2℃ and humidity within the range of 60±8%. The urgency is assigned a value of 0.5. The load characteristics of process C do not have strict requirements for immediate adjustment of temperature and humidity. It only needs to be maintained at 18-28℃ and 40-70%RH. The urgency is assigned a value of 0.2.

[0038] Energy consumption sensitivity: The load characteristics of process A show that a 5% reduction in air conditioning energy consumption will directly lead to a decrease in temperature control accuracy, exceeding the allowable range of the process, making it extremely sensitive to changes in energy consumption, with a sensitivity value of 0.9; the load characteristics of process B show that an 8% reduction in energy consumption can still maintain a normal thermal and humidity environment, with a sensitivity value of 0.5; the load characteristics of process C show that a 15% reduction in energy consumption will only have a slight impact on the environment, with a sensitivity value of 0.2. Based on the above quantification, the priority factors for the three types of processes are: process A (0.9, 0.8, 0.9), process B (0.6, 0.5, 0.5), and process C (0.3, 0.2, 0.2).

[0039] The dynamic parameter adjustment matrix is ​​input into the parallel computing module to synchronously calculate the operating status of multiple air conditioning modules and generate multiple sets of candidate operating parameters. The parallel computing module adopts a distributed computing architecture and performs parameter iteration calculations simultaneously based on the independent operating characteristics and interactive influence relationships of each air conditioning module. The dynamic parameter adjustment matrix needs to map the priority factors of each process with the air conditioning adjustment parameters (such as compressor frequency, fan speed, air supply temperature, and air supply humidity) to clarify the logic that "the higher the priority, the higher the priority of air conditioning parameter adjustment" and support real-time updates. Taking the automotive parts workshop as an example, the matrix rows represent process types (processes A, B, and C), the columns represent "priority factor dimensions + air conditioning adjustment parameters", and the matrix elements are the association weights (the higher the weight, the more sensitive the parameter adjustment is to the priority of the process): the association weight of "production criticality (0.9)" and "air supply temperature adjustment" of process A is set to 0.8, which means that when the production criticality of process A triggers the adjustment demand, the air supply temperature is adjusted first; its "heat and humidity demand urgency (0.8)" and "compressor frequency adjustment" association weight is set to 0.7 to ensure rapid temperature stabilization. For process B, the correlation weight between "urgency of heat and humidity demand (0.5)" and "supply air humidity adjustment" is set to 0.6, as the cleaning process is more sensitive to humidity; the correlation weight between "energy consumption sensitivity (0.5)" and "fan speed adjustment" is set to 0.5, balancing energy consumption and process requirements. For process C, the correlation weights between each priority factor and air conditioning parameter are all set to 0.3, with minor adjustments only made without affecting processes A and B. This matrix can automatically update the correlation weights based on real-time process load changes (e.g., an increase in the number of processing batches in process A, raising the production criticality to 0.95), generating a real-time adjustable parameter adjustment logic framework.

[0040] After constructing the dynamic parameter adjustment matrix, it needs to be input into the parallel computing module. The parallel computing module then synchronously calculates the operating status of multiple air conditioning modules, ultimately generating multiple sets of candidate operating parameters. The parallel computing module employs a distributed computing architecture (such as distributed nodes based on Hadoop). During the calculation process, the multiple air conditioning modules are first decomposed into independent computing nodes (compressor node, fan node, humidity control node). Then, considering the independent operating characteristics of each air conditioning module (e.g., compressor frequency only affects cooling capacity, fan speed only affects air delivery rate) and their interactive relationships (e.g., increasing fan speed increases energy consumption, potentially affecting the optimal value of compressor frequency), iterative parameter calculations are performed simultaneously to ensure computational efficiency and parameter adaptability.

[0041] Taking the air conditioning system (including 1 compressor, 2 fans, and 1 humidifier) ​​in an automotive parts workshop as an example: The parallel computing module allocates the correlation weights in the dynamic parameter adjustment matrix to the corresponding independent computing nodes. The compressor node calculates the candidate values ​​of compressor frequency (50Hz, 55Hz, 60Hz) based on the correlation weight of "urgency of heat and humidity demand in process A (0.8)" in the matrix. The fan node calculates the candidate values ​​of fan speed (1200r / min, 1400r / min, 1600r / min) based on the correlation weight of "humidity demand in process B". The humidifier node calculates the candidate values ​​of humidification capacity (2kg / h, 3kg / h, 4kg / h) based on the correlation weight of "humidity demand in process B". During the iterative calculation, each node synchronously interacts with data through distributed communication, fully considering the interaction and influence between modules: when the compressor frequency is set to 60Hz (strong cooling), the fan node will optimize the candidate speed value to 1600r / min according to the interaction logic of "strong cooling requires improving air supply efficiency to uniformly cool down", and the humidifier node will correspondingly reduce to 2kg / h to avoid excessively low humidity; if the compressor frequency is 50Hz (weak cooling), the fan node does not need a high speed to meet the air supply requirements, maintaining 1200r / min, and the humidifier node can maintain 3kg / h.

[0042] After multiple rounds of iterative optimization of parameter combinations, the parallel computing module ultimately outputs multiple sets of candidate operating parameters, each containing the operating parameters of all air conditioning modules. For the aforementioned workshop, the core candidate sets are three: Candidate Set 1: Compressor frequency 60Hz, fan speed 1600r / min, humidification capacity 2kg / h (prioritizing the high-criticality requirements of process A); Candidate Set 2: Compressor frequency 55Hz, fan speed 1400r / min, humidification capacity 3kg / h (balancing the requirements of processes A and B); Candidate Set 3: Compressor frequency 50Hz, fan speed 1200r / min, humidification capacity 4kg / h (prioritizing energy consumption control to meet the basic requirements of processes B and C).

[0043] Based on a preset target energy consumption threshold, candidate operating parameter sets are screened and optimized to obtain air conditioning operating parameter combinations under the target energy consumption. The screening process involves constructing a dual-objective evaluation function for energy consumption and process compliance, introducing parameter sensitivity analysis to predict the energy consumption impact of minor fluctuations in key adjustment parameters. A target energy consumption threshold is pre-set, combining the company's energy-saving goals and process energy consumption benchmarks. Taking an automotive parts workshop as an example, based on historical data and energy-saving requirements, the target energy consumption threshold for the workshop's air conditioning system is set at 120 kWh / h (10 hours of daily production, with an average daily energy consumption not exceeding 1200 kWh). To achieve the best balance between energy saving and process assurance, a dual-objective evaluation function that simultaneously considers "energy consumption compliance" and "process compliance" is constructed. The simplified formula for this function is: Evaluation score = (1 - actual energy consumption / target energy consumption threshold) × 0.5 + process satisfaction × 0.5 (full score 1.0, score ≥ 0.8 is qualified). The "process satisfaction" is calculated by comparing the deviation between the heat and humidity output corresponding to the air conditioning parameters and the process requirements (e.g., process A requires a temperature of 20±1℃, the parameter output temperature is 20.2℃, and the satisfaction is 0.98).

[0044] This function is used to evaluate and filter three candidate parameter sets. Specifically, Candidate Set 1: Actual energy consumption 135 kWh / h (exceeds the threshold), energy consumption compliance = -0.06, process compliance 0.95, evaluation score 0.415 (unacceptable, eliminated). Candidate Set 2: Actual energy consumption 118 kWh / h (below the threshold), energy consumption compliance ≈ 0.008, process compliance 0.92, evaluation score ≈ 0.468 (acceptable). Candidate Set 3: Actual energy consumption 105 kWh / h (below the threshold), energy consumption compliance = 0.0625, process compliance 0.85, evaluation score = 0.4875 (acceptable). After initial screening, Candidate Set 2 and Candidate Set 3 are retained.

[0045] To ensure the selected parameter combinations have good stability and robustness, parameter sensitivity analysis is introduced. This analysis aims to predict the potential impact of small fluctuations in key control parameters (such as compressor frequency and fan speed) on energy consumption and process compliance, preventing the system from deviating from the optimal state or violating constraints due to minor disturbances. Sensitivity analysis is performed using candidate set 2 (compressor frequency 55Hz, fan speed 1400r / min) as an example. Compressor frequency analysis: if the frequency increases from 55Hz to 56Hz (fluctuation of 1.8%), energy consumption increases to 122kWh / h (exceeding the threshold). Although the process compliance increases to 0.93, the risk of exceeding the energy consumption limit increases significantly. Therefore, the frequency must be limited to no more than 55Hz.

[0046] Analysis of fan speed revealed that if the speed decreased from 1400 r / min to 1350 r / min (fluctuation of 3.6%), energy consumption would drop to 115 kWh / h, and the process compliance rate would decrease to 0.89 (still acceptable). Therefore, the speed was optimized to 1350 r / min. Based on comprehensive evaluation and sensitivity analysis, the optimal combination of air conditioning operating parameters was determined to be: compressor frequency 55 Hz, fan speed 1350 r / min, and humidification capacity 3 kg / h. Under this combination, energy consumption was 115 kWh / h (below the threshold), and the process compliance rate was 0.89, achieving an effective balance between energy saving and process requirements.

[0047] S104 extracts peak energy consumption, load fluctuation rate, and process-energy consumption matching difference value according to industrial process scenarios, classifies the risk level of energy consumption exceeding the standard, and constructs a multi-dimensional feature matrix that includes process demand intensity, environmental interference coefficient, and equipment energy efficiency degradation information.

[0048] In one implementation, energy consumption peaks, load fluctuation rates, and process-energy consumption matching differences are extracted according to industrial process scenarios to generate energy consumption characteristic data. The energy consumption peak is the maximum energy consumption of air conditioning per unit time under the corresponding process scenario. The "unit time" for statistics needs to be determined according to the process operation cycle to ensure that the peak accurately reflects the highest energy consumption demand under the scenario. Taking a plastic injection molding workshop as an example, the process operation cycle of this workshop is 1 hour (including injection heating, mold cooling, product removal, etc.). When extracting the peak energy consumption, the production period from 8:00 to 18:00 every day is first locked, and the air conditioning energy consumption is counted in 1-hour units: 8:00-9:00 energy consumption 95kWh, 9:00-10:00 energy consumption 105kWh, 10:00-11:00 energy consumption 110kWh, 11:00-12:00 energy consumption 108kWh... By comparing the energy consumption values ​​of each time period, the peak energy consumption of 110kWh from 10:00 to 11:00 is determined to be the peak energy consumption of this process scenario on that day. Considering that there is a sudden increase in energy consumption in a short period of time during the injection heating stage, if the statistical period is adjusted to 15 minutes to further refine the data collection granularity, it may be possible to capture the instantaneous peak of 115kWh in the period from 10:30 to 10:45. The specific statistical unit needs to be flexibly determined based on the fluctuation pattern of process energy consumption. The core objective is to ensure that the extracted peak value can truly reflect the extreme energy consumption value under the process scenario.

[0049] Load fluctuation rate needs to consider both the amplitude and frequency of energy consumption changes per unit time. Amplitude reflects the "size" of energy consumption fluctuations, while frequency reflects the "intensity" of energy consumption fluctuations. Combining the two can quantify energy consumption stability and is an important component of energy consumption characteristic data. Taking a plastic injection molding workshop as an example, when extracting load fluctuation rate, the unit time is first set to 30 minutes. Amplitude calculation: If energy consumption increases from 90kWh to 110kWh within 30 minutes, the change amplitude is calculated using the formula "(current energy consumption - energy consumption of the previous period) / energy consumption of the previous period", and the result is (110-90) / 90≈22.2%; if it decreases from 105kWh to 100kWh, the change amplitude is (105-100) / 105≈4.8%. The amplitude value can be used to intuitively judge the strength of a single energy consumption fluctuation. Frequency Calculation: Count the number of times energy consumption changes by more than 5% within one hour. If there are three significant fluctuations within one hour (22.2% for the first 30 minutes, -8.1% for the next 30 minutes, and 10.5% for the following 30 minutes), the fluctuation frequency is 3 times / hour. The frequency value reflects the intensity of energy consumption fluctuations. Comprehensive Quantification: The load fluctuation rate is calculated using the method of "average amplitude × frequency". In the above scenario, the average amplitude is first calculated as (22.2% + 4.8% + 10.5%) / 3 ≈ 12.5%, and then multiplied by the fluctuation frequency of 3 times / hour, resulting in a load fluctuation rate of approximately 37.5% / hour. The higher this value, the more severe the fluctuation in air conditioning energy consumption and the worse the energy consumption stability under this process scenario.

[0050] The process-energy consumption matching difference value is obtained by comparing the deviation between the actual energy consumption of the air conditioner and the theoretical energy consumption required by the process. It is a key indicator for measuring the adaptability of energy consumption to process requirements and needs to be calculated in combination with the specific requirements of the plastic injection molding process. In a plastic injection molding workshop, when the process requires the mold temperature to be stable at 60℃, the theoretical energy consumption to meet the process requirements is calculated as 90kWh based on parameters such as air conditioning cooling efficiency (energy efficiency ratio 3.2) and workshop heat loss coefficient (15W / ℃), using the heat load calculation formula "Theoretical energy consumption = (target mold temperature - ambient reference temperature) × workshop heat loss coefficient / air conditioning energy efficiency ratio". If the actual collected air conditioning energy consumption is 110kWh, the process-energy consumption matching difference is calculated using the formula "(actual energy consumption - theoretical energy consumption) / theoretical energy consumption", resulting in (110-90) / 90≈22.2%. If the actual energy consumption is 85kWh, the difference is (85-90) / 90≈-5.6% (a negative deviation indicates that the energy consumption is lower than the theoretical value, requiring further judgment on whether the process requirements are met to avoid process non-compliance due to excessive energy saving). This indicator can clearly identify the imbalance between energy consumption and process requirements, providing direction for subsequent parameter optimization.

[0051] Through the above steps, the three major energy consumption characteristics data of plastic injection molding process scenario are extracted: peak energy consumption (115kWh / 15 minutes), load fluctuation rate (37.5% / hour), and process-energy consumption matching difference value (22.2%). This forms a complete energy consumption characteristic dataset, which lays the foundation for subsequent classification of energy consumption exceedance risk levels and construction of multi-dimensional feature matrix.

[0052] Based on the numerical ranges of peak energy consumption, load fluctuation rate, and process-energy consumption matching difference, energy consumption exceedance risk levels are classified, generating risk level identification data. Combining industry energy consumption standards, enterprise energy-saving targets, and process requirements, differentiated threshold ranges are set for three process scenarios: "plastic injection molding," "machining," and "parts cleaning," for peak energy consumption, load fluctuation rate, and process-energy consumption matching difference. This ensures that the thresholds accurately match the energy consumption patterns of each scenario: For peak energy consumption thresholds (based on 15-minute intervals), for plastic injection molding workshops: low risk (≤100kWh / 15 minutes), medium risk (101-120kWh / 15 minutes), and high risk (>120kWh / 15 minutes). For machining workshops: low risk (≤90kWh / 15 minutes), medium risk (91-110kWh / 15 minutes), and high risk (>110kWh / 15 minutes). Parts cleaning workshop: low risk (≤80kWh / 15 minutes), medium risk (81-100kWh / 15 minutes), high risk (>100kWh / 15 minutes).

[0053] For load fluctuation rate thresholds: Plastic injection molding workshop: low risk (≤20% / hour), medium risk (21-40% / hour), high risk (>40% / hour). Machining workshop: low risk (≤25% / hour), medium risk (26-35% / hour), high risk (>35% / hour). Parts cleaning workshop: low risk (≤30% / hour), medium risk (31-35% / hour), high risk (>35% / hour).

[0054] For the process-energy consumption matching difference threshold, the risk levels are as follows: Plastic injection molding workshop: low risk (≤±10%), medium risk (±11-±25%), high risk (>±25%). Machining workshop: low risk (≤±15%), medium risk (±16-±30%), high risk (>±30%). Parts cleaning workshop: low risk (≤±15%), medium risk (±16-±30%), high risk (>±30%).

[0055] The risk level is divided by a weighted voting method, which assigns different weights to the three energy consumption characteristics (peak energy consumption weight 0.4, load fluctuation rate weight 0.3, and process-energy consumption matching difference value weight 0.3). The final level is determined by summing the weights of the risk levels to which each characteristic belongs. Taking the energy consumption characteristic data of a plastic injection molding workshop (peak energy consumption 115kWh / 15 minutes, load fluctuation rate 37.5% / hour, process-energy consumption matching difference value 22.2%) as an example, the peak energy consumption of 115kWh / 15 minutes belongs to medium risk, with a weight contribution of 0.4 × "medium risk" (recorded as 2 points) = 0.8 points; the load fluctuation rate of 37.5% / hour belongs to medium risk, with a weight contribution of 0.3 × 2 points = 0.6 points; the process-energy consumption matching difference value of 22.2% belongs to medium risk, with a weight contribution of 0.3 × 2 points = 0.6 points; the total weight score is 0.8 + 0.6 + 0.6 = 2 points, corresponding to the medium risk level, and the risk level label data generated is "plastic injection molding workshop - medium risk".

[0056] Information on process demand intensity, environmental interference coefficient, and equipment energy efficiency degradation in industrial process scenarios is collected to construct a multi-dimensional feature matrix incorporating these three metrics. For each of the three core feature dimensions—process demand intensity, environmental interference coefficient, and equipment energy efficiency degradation—implementable quantification rules are developed to ensure that the collected data can be directly used for matrix construction. Process demand intensity (quantification range 0-1): This value considers both the precision requirements for temperature and humidity and the duration of the demand. Higher precision requirements (e.g., temperature control accuracy ±1℃ better than ±3℃) and longer durations (e.g., 10 hours per day better than 8 hours) result in a higher intensity value. For example, in a plastic injection molding workshop: the mold temperature needs to be stable at 60±1℃ (high precision), lasting 10 hours per day (long duration), resulting in a process demand intensity quantified as 0.9.

[0057] Machining workshop: Temperature requirement 20±3℃ (medium precision), 8 hours daily (medium duration), intensity quantified as 0.6. Parts cleaning workshop: Humidity requirement 60±8% (low precision), 6 hours daily (short duration), intensity quantified as 0.4. Environmental interference coefficient (quantification range 0-1): Evaluated based on the actual intensity of interference sources such as electromagnetic interference, dust, and vibration in the industrial environment. The more types and stronger the interference sources, the higher the coefficient. For example, plastic injection molding workshop: Injection molding machine operation generates strong electromagnetic interference, no significant dust but equipment vibration exists, environmental interference coefficient quantified as 0.8. Machining workshop: Machine tool operation brings strong vibration, moderate electromagnetic interference, no obvious dust, coefficient quantified as 0.7; Parts cleaning workshop: Only slight water flow noise exists, no electromagnetic interference or dust, coefficient quantified as 0.3.

[0058] Equipment energy efficiency degradation information (quantification range 0-1): The energy efficiency degradation rate is calculated by combining the service life of the air conditioning equipment and the maintenance frequency. The longer the service life and the less timely the maintenance, the higher the degradation rate (i.e., the larger the quantification value). The calculation formula can be simplified to: Degradation rate = (service life × 0.1) - (average number of maintenance times per year × 0.05), and the result is truncated to the 0-1 range. For example: Plastic injection molding workshop: Air conditioning used for 5 years, average maintenance 2 times per year, degradation rate = (5 × 0.1) - (2 × 0.05) = 0.5 - 0.1 = 0.4. Machining workshop: Air conditioning used for 8 years, average maintenance 1 time per year, degradation rate = (8 × 0.1) - (1 × 0.05) = 0.8 - 0.05 = 0.75; Parts cleaning workshop: Air conditioning used for 2 years, average maintenance 4 times per year, degradation rate = (2 × 0.1) - (4 × 0.05) = 0.2 - 0.2 = 0.

[0059] The matrix uses row vectors to represent different industrial process scenarios, and column vectors to correspond to "process demand intensity," "environmental interference coefficient," and "equipment energy efficiency degradation information," respectively. Each matrix element is a specific quantitative value of the feature. Combining the feature data of the above three types of process scenarios, a multi-dimensional feature matrix is ​​constructed as follows: Row 1 (plastic injection molding workshop): [0.9, 0.8, 0.3]; Row 2 (machining workshop): [0.6, 0.7, 0.6]; Row 3 (parts cleaning workshop): [0.4, 0.3, 0.1]. This matrix clearly presents the differences of each feature under different scenarios. For example, the plastic injection molding workshop has "high process demand intensity (0.9) + large environmental interference (0.8)", and the machining workshop has "severe equipment energy efficiency degradation (0.6)", providing feature support for subsequent targeted energy-saving strategies. Furthermore, if the characteristic data of a newly added process scenario (such as an automotive painting workshop) is "process demand intensity 0.8, environmental interference coefficient 0.9, and equipment energy efficiency degradation information 0.2", it can be directly added to the matrix as row 4 to expand the scenario coverage of the matrix and provide complete characteristic data support for the subsequent gradient boosting tree algorithm to select key factors.

[0060] S105 compares real-time energy consumption with historical data from the same process and period to identify abnormal signals such as short-term energy consumption spikes and load-energy consumption mismatch. It then uses the slope of the energy consumption-process parameter correlation curve to generate a dynamic correction factor for energy consumption optimization.

[0061] In one implementation, real-time energy consumption data of the industrial process is retrieved and compared with historical energy consumption data of the same process during the same period to establish a data comparison dimension. Real-time energy consumption data needs to focus on the current process operation cycle and include two core data types: instantaneous energy consumption of the air conditioner and average energy consumption per unit time within the current process operation cycle, reflecting short-term fluctuations and overall trends in energy consumption, respectively. Instantaneous energy consumption refers to the immediate energy consumption value of the air conditioner at a specific moment, and short-term fluctuations in energy consumption need to be captured through high-frequency data acquisition. Taking the fabric dyeing process in a textile printing and dyeing workshop as an example, this process runs daily from 9:00 to 17:00 (8-hour cycle). Instantaneous energy consumption data is retrieved every 5 minutes through the air conditioner energy consumption monitoring terminal. Specific data include 120kWh at 9:00, 122kWh at 9:05, 118kWh at 9:10, and 115kWh at 17:00. This data can accurately present the instantaneous changes in energy consumption during the dyeing process (such as heating and heat preservation dyeing), avoiding the omission of key fluctuation information due to excessively long acquisition intervals. Average energy consumption per unit time refers to the average energy consumption of air conditioning within a certain period. It needs to be divided into time periods to reflect the phased trend of energy consumption. Taking this dyeing process as an example, the average energy consumption of each time period is calculated by dividing it into 1-hour units: In the 9:00-10:00 time period, the average energy consumption is calculated to be 119 kWh / h based on all instantaneous energy consumption data (a total of 12 data points) in this time period; the average energy consumption in the 10:00-11:00 time period is 121 kWh / h. By comparing the average values, the energy consumption differences of different dyeing stages (such as 10:00-11:00 is the concentrated dyeing period, and the energy consumption is slightly higher than that of the initial heating period) can be clearly identified, laying the foundation for subsequent comparison with historical data.

[0062] Historical data must match two core conditions: "same season, same production load," to avoid interference from seasonal environmental differences (such as higher cooling energy consumption in summer) and changes in production load (such as adjustments to process time due to increases or decreases in order volume). It must also include both average and fluctuation range indicators. Taking the dyeing process in a textile printing and dyeing workshop as an example, energy consumption data from the same period last year (e.g., May 2023) under the same production load (5000 meters of fabric dyed daily, consistent with the current level) should be retrieved. Average energy consumption: The average air conditioning energy consumption for the dyeing process from 9:00 to 17:00 daily in May last year was 118 kWh / h; Fluctuation range: The highest energy consumption in the same period last year was 125 kWh / h, and the lowest was 112 kWh / h, with a fluctuation range of 112-125 kWh / h.

[0063] Based on preset energy consumption difference thresholds and load-energy consumption matching benchmarks, real-time energy consumption is compared with historical data from the same process and time period to identify abnormal signals such as short-term energy consumption spikes and load-energy consumption imbalances. A two-way comparison dimension is established between real-time energy consumption and historical data from the same process and time period, focusing on the three core attributes of energy consumption data: instantaneous fluctuation, time-period average, and overall range. This ensures the comprehensiveness and relevance of the data comparison. For real-time instantaneous energy consumption versus historical average instantaneous energy consumption during the same period: instantaneous energy consumption data within the current process cycle (e.g., 122 kWh at 9:05) is selected and compared with the historical average instantaneous energy consumption (117 kWh) for the same process and time period (9:00-9:10) during the same period (e.g., May of last year). This focuses on capturing short-term abnormal instantaneous fluctuations in energy consumption and avoids misjudgments due to the randomness of data from a single point in time.

[0064] Regarding real-time average energy consumption per unit time versus historical average energy consumption per unit time: Calculate the average energy consumption of the current process at various time periods (e.g., 9:00-10:00) (119 kWh / h), and compare it with the average energy consumption of the same process at the same time period in the past (118 kWh / h). Analyze whether the phased trend of energy consumption deviates from historical norms and reflect whether the overall change in energy consumption is reasonable. Regarding real-time energy consumption fluctuation range versus historical energy consumption fluctuation range: Statistically calculate the highest and lowest energy consumption values ​​(115-122 kWh) within the current process operating cycle, and compare them with the energy consumption fluctuation range of the same process in the past (112-125 kWh). Determine whether the overall stability of the current energy consumption conforms to historical patterns and investigate abnormal risks caused by large-scale fluctuations.

[0065] Based on the enterprise's energy-saving goals and the process requirements for energy consumption stability, two core judgment criteria are formulated to provide a quantitative basis for anomaly identification. The energy consumption difference threshold includes the following information: Short-term energy consumption increase threshold: For rapid changes in instantaneous energy consumption, a threshold of 8% is set for energy consumption increase within 10 minutes. That is, when the real-time energy consumption increases by more than 8% compared to the previous moment within 10 minutes, it is judged as a short-term energy consumption surge. Time-period average deviation threshold: For the overall deviation of the time-period average energy consumption, a threshold of ±5% is set for the deviation of the real-time unit time average energy consumption from the historical average for the same period. Exceeding this range is considered an abnormal energy consumption average, requiring further investigation. The load-energy consumption matching benchmark is as follows: Based on the inherent correlation between process load and energy consumption, a benchmark range for the correlation coefficient is set. Taking the dyeing process in a textile printing and dyeing workshop as an example, when the process load (fabric dyeing volume) increases by 10%, the air conditioning energy consumption should increase by 5%-7% simultaneously. This coefficient range is the load-energy consumption matching benchmark. If the actual energy consumption increase exceeds or falls below this range, it is judged as a load-energy consumption imbalance.

[0066] The system identifies short-term energy consumption spikes by monitoring whether the increase in real-time energy consumption per unit time (e.g., 10 minutes) exceeds a preset threshold. In a textile dyeing workshop, the instantaneous energy consumption was 118 kWh at 9:10 AM and rose to 128 kWh at 9:20 AM. The increase within 10 minutes was calculated as (128-118) / 118 ≈ 8.47%, exceeding the preset threshold of 8%. Therefore, it was identified as a "short-term energy consumption spike" anomaly. Investigation revealed that the spike was caused by the addition of a heating device to the dyeing process, leading to a sudden increase in the workshop's heat load. The air conditioning system needed to increase its cooling capacity to maintain the required temperature and humidity, necessitating further optimization of the equipment's operation.

[0067] The identification of load-energy consumption mismatch anomaly signals is achieved by comparing the correlation between actual energy consumption changes and process load changes to see if it conforms to a preset benchmark. In the textile printing and dyeing workshop, during the period from 10:00 to 11:00, the process load (amount of dyed fabric) increased by 10% compared to the previous period (from 500 meters / hour to 550 meters / hour). According to the load-energy consumption matching benchmark, air conditioning energy consumption should increase by 5%-7%; however, the actual energy consumption increased from 119 kWh / h to 129 kWh / h, an increase of (129-119) / 119≈8.4%, exceeding the benchmark upper limit of 7%. Therefore, it was determined to be an "load-energy consumption mismatch" anomaly signal. Upon inspection, the imbalance was found to be caused by the aging of the air conditioning fan. When the load increased, the fan needed to operate beyond its rated power, resulting in an energy consumption increase higher than the process load increase. Timely maintenance or replacement of the fan is necessary to restore the matching relationship.

[0068] Energy consumption data and process parameters for the corresponding industrial processes are collected, and an energy consumption-process parameter correlation curve is constructed. The slope of the curve at the current process parameter node is calculated, and this slope is used to generate a dynamic correction factor for energy consumption optimization. For the dyeing process in the textile printing and dyeing workshop, the core process parameter with a significant impact on energy consumption—"dyeing temperature (°C)"—is prioritized. Multiple sets of "dyeing temperature-air conditioning energy consumption" data are simultaneously collected through energy consumption monitoring terminals and process parameter sensors to ensure that the data covers the range of commonly used process parameters. Specific data are as follows: At a dyeing temperature of 60°C, air conditioning energy consumption is 115 kWh / h; at a dyeing temperature of 65°C, air conditioning energy consumption is 122 kWh / h; at a dyeing temperature of 70°C, air conditioning energy consumption is 130 kWh / h; and at a dyeing temperature of 75°C, air conditioning energy consumption is 138 kWh / h.

[0069] Using the core process parameter "dyeing temperature" as the X-axis and the corresponding air conditioning energy consumption as the Y-axis, the above multiple sets of data points were imported into data analysis tools (such as Excel and Matlab) for fitting, generating a linear energy consumption-process parameter correlation curve. Through linear regression calculation, the curve equation was obtained as y = 1.4x + 21 (where x is the dyeing temperature and y is the air conditioning energy consumption). The curve trend clearly shows the correlation that "for every 5°C increase in dyeing temperature, air conditioning energy consumption increases by approximately 7 kWh / h." This curve accurately reflects the intrinsic relationship between changes in dyeing temperature and energy consumption, laying the foundation for subsequent slope calculations.

[0070] The slope of the curve at the current process parameter node, and the slope of the energy consumption-process parameter correlation curve, directly reflect the degree of influence of process parameter changes on energy consumption. The larger the slope, the more sensitive the energy consumption is to changes in the process parameter. For the linear curve y=1.4x+21 constructed in this study, its slope is a constant value of 1.4. If the constructed curve is a nonlinear curve (such as a quadratic function y=ax²+bx+c), then the slope of the tangent line at the current process parameter node needs to be calculated using the derivative formula to reflect the instantaneous impact of parameter changes on energy consumption at that node. Assuming the actual operating parameter of the current dyeing process is "dyeing temperature 70℃", the slope of the corresponding curve node is 1.4. This value means that under the current process conditions, for every 1℃ change in dyeing temperature, the air conditioning energy consumption will change in the same direction by 1.4 kWh / h (e.g., a 1℃ increase in temperature increases energy consumption by 1.4 kWh / h; a 1℃ decrease in temperature decreases energy consumption by 1.4 kWh / h).

[0071] A dynamic correction factor for energy consumption optimization is generated using the slope. The calculation of this correction factor requires consideration of the slope, the preset energy consumption optimization target, and the adjustable range of process parameters to ensure that the correction scheme meets energy-saving requirements without violating process constraints. The calculation formula is: Correction Factor = (Target Energy Consumption Reduction × Current Energy Consumption) / (Slope × Adjustable Range of Process Parameters). Taking the current scenario of a textile printing and dyeing workshop as an example, the target energy consumption reduction is set at 5% based on the company's energy-saving goals; the current dyeing temperature of 70℃ corresponds to an air conditioning energy consumption of 130 kWh / h; the slope is 1.4 (already calculated). Because the dyeing process requires a stable temperature between 67-73℃, the adjustable range of the dyeing temperature is ±3℃. Substituting the above parameters into the formula, the correction factor is calculated as follows: Correction Factor = (5% × 130) / (1.4 × 3) ≈ 6.5 / 4.2 ≈ 1.55. This correction factor means that to achieve a 5% reduction in energy consumption (i.e., energy consumption decreasing from 130 kWh / h to 123.5 kWh / h), at the current dyeing temperature of 70℃, the energy consumption optimization target can be achieved by lowering the dyeing temperature by approximately 1.55℃ (within an adjustable range of ±3℃), combined with synchronous adjustments to the air conditioning fan parameters. Furthermore, this correction factor has a dynamic update capability—if subsequent process parameters change (e.g., the dyeing temperature rises to 75℃, and the current energy consumption becomes 138 kWh / h), a new correction factor can be recalculated using the formula, ensuring that it always adapts to the real-time process conditions.

[0072] S106 groups users based on the production scale, process complexity, and equipment configuration of industrial enterprises. It uses the gradient boosting tree algorithm to screen key influencing factors of the multidimensional feature matrix and energy consumption optimization dynamic correction factor, and integrates energy consumption optimization parameters, process constraints, and equipment operating limit information to construct a personalized energy-saving control model.

[0073] In one implementation, quantifiable grouping criteria are established based on three core dimensions: "production scale, process complexity, and equipment configuration," avoiding biases caused by subjective classification. Production scale: Based on the enterprise's average daily industrial output value (in ten thousand yuan), it is divided into small-scale (average daily output value ≤ 500,000 yuan), medium-scale (500,000 yuan < average daily output value ≤ 2 million yuan), and large-scale (average daily output value > 2 million yuan). Process complexity: Based on a comprehensive score (out of 10 points) of the number of process steps (e.g., independent steps such as raw material pretreatment, processing, and testing) and parameter control accuracy (e.g., temperature control accuracy ±1℃ is high, ±3℃ is medium, and ±5℃ is low), it is divided into low complexity (score ≤ 4 points), medium complexity (score ≤ 4 points), and large complexity (score ≤ 4 points). Score < 7 points ≤ High complexity (score > 7 points); Equipment configuration: Based on the number of modules in the air conditioning system (such as the total number of independent modules such as compressor, fan, humidifier, etc.) and the service life of the equipment (annual average decay rate < 5% is new, 5%-10% is medium, > 10% is old), it is divided into basic configuration (number of modules ≤ 5 and service life of equipment > 10%), standard configuration (5 < number of modules ≤ 10 and 5% ≤ service life of equipment ≤ 10%), and high-end configuration (number of modules > 10 and service life of equipment < 5%).

[0074] Taking four companies in the machinery manufacturing industry as examples, and grouping them according to the above dimensions: Company A: Daily output value of 300,000 yuan (small scale), 3 process steps with temperature control accuracy of ±5℃ (score of 3 points, low complexity), 4 air conditioning modules with equipment service life of 12% (basic configuration) → Grouped as "Small Scale - Low Complexity - Basic Configuration Group"; Company B: Daily output value of 1.2 million yuan (medium scale), 5 process steps with temperature control accuracy of ±3℃ (score of 6 points, medium complexity), 8 air conditioning modules with equipment service life of 8% (standard configuration) → Grouped as "Medium Scale - Medium Complexity Group" "Complexity - Standard Configuration Group"; Company C: Daily output value of 2.8 million yuan (large scale), 8 process steps with temperature control accuracy of ±1℃ (score of 9 points, high complexity), 15 air conditioning modules with equipment service life of 3% (high-end configuration) → grouped as "Large Scale - High Complexity - High-end Configuration Group"; Company D: Daily output value of 600,000 yuan (medium scale), 4 process steps with temperature control accuracy of ±3℃ (score of 5 points, medium complexity), 6 air conditioning modules with equipment service life of 7% (standard configuration) → grouped with Company B, classified as "Medium Scale - Medium Complexity - Standard Configuration Group".

[0075] The gradient boosting tree algorithm iteratively generates multiple decision trees and accumulates the prediction results of each tree. This quantifies the impact weight of each feature on energy consumption optimization, thereby selecting core factors, reducing redundant feature interference, and improving the accuracy and computational efficiency of subsequent models. Taking a "medium-scale, medium-complexity, standard configuration group" (such as companies B and D) as an example, the input feature set contains two types of data: a multi-dimensional feature matrix: the row vectors represent the company's process scenarios (such as machining and parts cleaning), and the column vectors represent the process demand intensity (0-1 quantization), environmental interference coefficient (0-1 quantization), and equipment energy efficiency degradation information (0-1 quantization). For example, the feature vector for company B's machining scenario is [0.7, 0.6, 0.5]. The dynamic correction factor for energy consumption optimization is a correction factor calculated based on the slope of the correlation curve between energy consumption and process parameters for this group of companies. For example, the correction factor for company B's machining scenario is 1.2, and for the parts cleaning scenario it is 0.8.

[0076] Initialize the model: Use the mean of all input features as the initial predicted value and calculate the initial residual (the difference between the actual energy consumption optimization effect and the initial predicted value); Iteratively generate decision trees: In each iteration, build a decision tree to fit the residual. For example, if the first round of trees finds that the residual is significantly reduced when "process demand intensity > 0.6", the second round of trees captures the impact of "equipment energy efficiency degradation information > 0.4" on the residual. Iterate to generate 100 decision trees (the number of trees can be adjusted according to the amount of data); Calculate feature importance: Calculate the importance score by the contribution of the feature to the node split in all decision trees (such as the reduction of the Gini coefficient). The higher the score, the greater the impact on energy consumption optimization.

[0077] For the "Medium-Scale - Medium Complexity - Standard Configuration Group," the top three key factors by importance score were selected: Process Demand Intensity: 0.35 (a core influencing factor; the higher the process demand, the more crucial it is for air conditioning to meet process requirements, resulting in stronger constraints on energy consumption optimization); Equipment Energy Efficiency Degradation Information: 0.28 (the more severe the equipment degradation, the smaller the space for energy consumption optimization, requiring targeted parameter adjustments to balance energy efficiency and process demands); Energy Consumption Optimization Dynamic Correction Factor: 0.22 (directly reflects the potential for energy consumption optimization under current process parameters; the larger the correction factor, the more significant the optimization space). The "Environmental Interference Coefficient" scored only 0.15. Because the environmental interference in this group of enterprises is relatively stable (e.g., small fluctuations in electromagnetic interference in machining workshops), it was determined to be a non-critical factor and was removed.

[0078] Based on the selected key influencing factors, three types of key information are integrated to establish a "factor-parameter-constraint" correlation logic, and a personalized model adapted to specific user groups is constructed to ensure that the model meets both energy-saving goals and the actual limitations of processes and equipment. Taking a machining scenario of "medium scale-medium complexity-standard configuration group" as an example: Energy consumption optimization parameters: Based on the optimization target parameters determined by key factors, such as the target energy consumption threshold (110kWh / h) and the energy consumption reduction target (5%), are jointly determined by "process demand intensity" and "correction factor". When the demand intensity is high, the reduction target is appropriately relaxed.

[0079] Process constraints: The rigid environmental requirements of the process, such as the workshop temperature needing to be stable at 20±2℃ and relative humidity at 50±5% during machining. Exceeding these limits will affect machining accuracy. The strictness of the constraints is directly determined by the "intensity of process requirements". Equipment operating limit information: The physical operating boundaries of air conditioning equipment, such as the compressor frequency upper limit of 60Hz and lower limit of 30Hz, and the fan speed upper limit of 1800r / min and lower limit of 800r / min. These are corrected by the "equipment energy efficiency degradation information". When the degradation is severe, the upper limit is appropriately reduced (e.g., the upper limit of the compressor of old equipment is set to 55Hz).

[0080] Using "minimizing energy consumption" as the objective function and process constraints and equipment limits as boundary conditions, the correlation equation between key factors and air conditioning operating parameters is established: Objective function: Energy consumption = 0.35 × process demand intensity × compressor frequency + 0.28 × equipment energy efficiency degradation information × fan speed - 0.22 × correction factor × humidification capacity (coefficients adjusted based on feature importance score), with the goal of making energy consumption ≤ 110 kWh / h; Constraints: 20℃ ≤ workshop temperature ≤ 22℃ (corresponding to compressor frequency 35-55Hz), 45%RH ≤ relative humidity ≤55% (corresponding to humidification capacity of 2-4 kg / h), 30Hz≤compressor frequency≤55Hz, 800r / min≤fan speed≤1800r / min; Parameter solution: Solve the equation using the gradient descent algorithm to obtain the optimal air conditioning operating parameters that meet the constraints. For example, when the process demand intensity is 0.7, the equipment energy efficiency degradation is 0.5, and the correction factor is 1.2, the optimal parameters are compressor frequency 45Hz, fan speed 1200r / min, and humidification capacity 3 kg / h. At this time, the energy consumption is about 108 kWh / h, which meets the target and constraints.

[0081] The model parameters differ for different user groups. For example, for the "Large-Scale - High Complexity - High-End Configuration Group" (Enterprise C), the key factors are "Process Demand Intensity (0.4), Correction Factor (0.3), and Equipment Energy Efficiency Decay (0.2)" (due to the newness of the equipment, the decay effect is reduced); the process constraints are more stringent (temperature 20±1℃), and the equipment limits are higher (compressor frequency upper limit 65Hz); the objective function is adjusted to energy consumption = 0.4 × process demand intensity × compressor frequency + 0.3 × correction factor × fan speed - 0.2 × equipment energy efficiency decay × fresh air volume, to achieve personalized adaptation.

[0082] S107 generates real-time dynamic adjustment commands and time-sharing energy-saving control strategies based on the target operating parameters output by the personalized energy-saving control model, combined with the time correlation information of industrial process timing and air conditioning equipment operating characteristics.

[0083] In one implementation, target operating parameters output by a personalized energy-saving control model are extracted, and the industrial process timing and air conditioning equipment operating characteristics are collected simultaneously to establish a time-related mapping relationship among the three, generating parameter-time-equipment correlation data. To generate this data, the target operating parameters output by the personalized energy-saving control model must first be extracted. This model outputs adapted air conditioning operating parameters based on the characteristics of different user groups, and the parameters must cover the core adjustment dimensions of the air conditioning system to ensure that process requirements and energy-saving goals are met. Taking a "medium-scale - medium-complexity - standard configuration group" machining workshop as an example, the target operating parameters extracted from the model include: compressor frequency 45Hz (maintaining workshop temperature 20±2℃), fan speed 1200r / min (ensuring air circulation efficiency), humidification capacity 3kg / h (controlling relative humidity 50±5%), and fresh air volume 200m³ / h (balancing indoor air quality and energy consumption).

[0084] After extracting the target operating parameters, it is necessary to simultaneously collect data on the industrial process sequence and the operating characteristics of the air conditioning equipment. Specifically, regarding the industrial process sequence: clarify the time nodes, duration, and corresponding demand intensity of each process step in the machining workshop, specifically 8:00-10:00 (rough machining of engine cylinder block, process demand intensity 0.9, requiring strict temperature control), 10:00-12:00 (semi-finishing of parts, demand intensity 0.7), 13:00-15:00 (precision testing, demand intensity 0.8, requiring higher temperature control accuracy), and 15:00-17:00 (finished product assembly, demand intensity 0.5). Regarding the operating characteristics of the air conditioning equipment: record the equipment's operating performance under different time periods and parameters. For example, from 8:00-10:00 (peak grid load period), voltage fluctuations are likely to occur when the compressor frequency exceeds 50Hz; from 15:00-17:00 (lower ambient temperature), the fan speed below 1000r / min can still meet the air circulation requirements.

[0085] Finally, a time-series mapping relationship is established among the three elements to generate parameter-time-equipment correlation data. Using time as the link, the target operating parameters are associated with the corresponding process time sequence and equipment characteristics one by one, forming time-period correlation data. For example, 8:00-10:00 (cylinder rough machining): the correlation data is "compressor frequency 45Hz + process requirement intensity 0.9 + equipment power grid peak frequency ≤ 50Hz"; 15:00-17:00 (finished product assembly): the correlation data is "fan speed 1200r / min + process requirement intensity 0.5 + equipment speed can be reduced to 1000r / min at low ambient temperatures"; by sequentially completing the correlation for all time periods, a complete parameter-time-equipment correlation data sequence is finally generated, clearly presenting the adaptation relationship among the three elements at different time nodes.

[0086] Based on parameter-time-equipment correlation data, and targeting the real-time operating status of industrial processes, the deviation between the target operating parameters and the actual operating parameters is calculated in real time. Combined with the real-time operating characteristics of the air conditioning equipment, real-time dynamic adjustment instructions are generated. These instructions include the direction of parameter adjustment, the adjustment range, and the execution time. The frequency of instruction generation is synchronized with the frequency of process status changes.

[0087] Clearly defining the industrial process sequence and the operating characteristics of air conditioning equipment lays the foundation for subsequent data construction. Regarding the industrial process sequence, the time nodes, durations, and corresponding demand intensities for each process stage are clearly defined. Taking a machining workshop as an example, the sequence is: 8:00-10:00 (rough machining of engine cylinder blocks, process demand intensity 0.9, requiring strict temperature control), 10:00-12:00 (semi-finishing of parts, demand intensity 0.7), 13:00-15:00 (precision testing, demand intensity 0.8, requiring higher temperature control accuracy), and 15:00-17:00 (finished product assembly, demand intensity 0.5). Regarding the operating characteristics of the air conditioning equipment, its performance under different time periods and parameters needs to be recorded. For example, in this workshop, voltage fluctuations are likely to occur when the compressor frequency exceeds 50Hz during 8:00-10:00 (peak grid load period), while the air circulation requirement can still be met when the fan speed is below 1000 r / min during 15:00-17:00 (lower ambient temperature).

[0088] Next, using time as a link, the "target operating parameters - process sequence - equipment characteristics" are correlated to form parameter-sequence-equipment correlation data for each time period. 8:00-10:00 (Cylinder block rough machining): Corresponding data is "Compressor frequency 45Hz + process demand intensity 0.9 + equipment power grid peak frequency ≤ 50Hz". 10:00-12:00 (Parts semi-finishing): Corresponding data is "Compressor frequency 43Hz + process demand intensity 0.7 + equipment frequency can be adjusted between 40-50Hz during normal periods". 13:00-15:00 (Precision testing): Corresponding data is "Compressor frequency 46Hz + process demand intensity 0.8 + equipment power grid off-peak frequency can be adjusted between 42-52Hz". 15:00-17:00 (finished product assembly): The associated data is "fan speed 1200r / min + process requirement intensity 0.5 + equipment speed can be reduced to 1000r / min under low ambient temperature"; the association of all time periods is completed in sequence, and finally a complete parameter-time series-equipment association data sequence is generated, clearly showing the adaptation relationship of the three at different time nodes.

[0089] Subsequently, based on parameter-time-equipment correlation data, and considering the real-time operating status of the industrial process, real-time dynamic adjustment commands are generated by calculating the deviation between the target operating parameters and the actual operating parameters, combined with the real-time operating characteristics of the air conditioning equipment. For example, taking the operating data of a machining workshop at 8:30 (corresponding to the 8:00-10:00 time period, with a target compressor frequency of 45Hz in the correlation data), the actual values ​​of the core air conditioning operating parameters are collected in real time and compared with the target values. The calculations show: target compressor frequency 45Hz, actual frequency 48Hz, deviation value +3Hz (actual is higher than target); target relative humidity 50%, actual humidity 55%, deviation value +5%RH (actual is higher than target); target fresh air volume 200m³ / h, actual fresh air volume 190m³ / h, deviation value -10m³ / h (actual is lower than target). (Target); Determine the feasibility of correction based on equipment characteristics: Refer to the air conditioning equipment operating characteristics in the parameter-time series-equipment related data (compressor frequency ≤ 50Hz during peak grid load period 8:00-10:00), and determine whether the deviation can be corrected by adjustment and the safe correction range—the actual compressor frequency is 48Hz, which is higher than the target of 45Hz, but does not exceed the 50Hz limit, so it can be adjusted by reducing the frequency; the actual humidity is 55%, the equipment humidification module is currently operating normally, so it can be corrected by reducing the humidification amount; the actual fresh air volume is 190m³ / h, the equipment fresh air valve is not faulty, so the fresh air volume can be increased by increasing the valve opening;

[0090] Real-time dynamic adjustment commands are generated. Specifically, these commands include the direction, magnitude, and execution time of parameter adjustments, and the frequency of command generation is synchronized with the frequency of process status changes (e.g., if the process fluctuates every 10 minutes, a command is generated every 10 minutes). To address the aforementioned deviations, the generated adjustment commands are as follows: "Compressor frequency: Reduce from 48Hz to 45Hz, direction: decrease, magnitude: -3Hz, execution time: within 5 minutes, to avoid temperature fluctuations caused by a sudden frequency drop"; "Humidification: Reduce from 3kg / h to 2kg / h, direction: decrease, magnitude: -1kg / h, execution time: within 3 minutes, to ensure humidity gradually decreases to 50%"; "Fresh air volume: Increase from 190m³ / h to 200m³ / h, direction: increase, magnitude: +10m³ / h, execution time: within 2 minutes, to avoid sudden changes in fresh air volume affecting indoor temperature."

[0091] To generate a time-segmented energy-saving control strategy, it is necessary to first clarify the process requirements for different time periods based on the time-segmentation in the industrial process sequence. Then, combining the intensity of process requirements for each time period with the differences in the operating characteristics of air conditioning equipment at different times, the target operating parameters are adapted and optimized. Finally, the core control rules for each time period are clarified. Taking a machining workshop as an example: First, the industrial process sequence of the machining workshop is sorted out, and the daily production cycle is divided into multiple core time periods according to the differences in process requirements. Each time period corresponds to a clear process content and demand intensity. High demand period: 8:00-10:00 (rough machining of engine cylinder block, process demand intensity 0.9, strict temperature control is required to ensure machining accuracy), 13:00-15:00 (precision testing, process demand intensity 0.8, higher temperature control accuracy requirements). Medium demand period: 10:00-12:00 (semi-finishing of parts, process demand intensity 0.7, less stringent temperature and humidity requirements than rough machining). Low demand period: 15:00-17:00 (finished product assembly, process demand intensity 0.5, no strict temperature and humidity accuracy requirements).

[0092] Based on the intensity of process demand and equipment characteristics, the target operating parameters for each time period were adjusted to suit specific time periods. For each core time period, the intensity of process demand and the operating characteristics of the air conditioning equipment were analyzed, and the target operating parameters were adjusted accordingly. High demand period (8:00-10:00): Process demand intensity 0.9 (high temperature control requirements). The air conditioning equipment is operating during peak grid load (compressor frequency exceeding 50Hz is prone to voltage fluctuations). Therefore, the target compressor frequency was slightly adjusted from the base value of 45Hz to 46Hz (ensuring temperature stability at 20±2℃). The fresh air volume was maintained at 200m³ / h to balance air quality, while limiting the compressor frequency to no more than 50Hz to avoid grid fluctuations affecting equipment operation. Medium demand period (10:00-12:00): Process demand intensity 0.7. The air conditioning equipment is operating during off-peak grid periods (greater frequency adjustment range). The target compressor frequency was adjusted to 43Hz, the fan speed was maintained at 1200r / min, and the humidification capacity was 2.5kg / h, moderately reducing energy consumption while meeting process requirements. During periods of low demand (15:00-17:00): the process demand intensity is 0.5, the ambient temperature is low (air humidity is easy to maintain), and the air conditioning equipment fan can still meet the air circulation requirements at low speed. Therefore, the target fan speed is reduced from 1200r / min to 1000r / min, and the humidification capacity is reduced from 3kg / h to 1kg / h to reduce unnecessary energy consumption.

[0093] The time-segmented energy-saving control strategy needs to clearly define the core operating parameter ranges, module start-stop rules, and energy consumption control benchmarks for each time period to ensure the strategy can be implemented effectively. For 8:00-10:00 (high demand): the core parameter range is compressor frequency 45-50Hz, fan speed 1200-1300r / min, and humidification capacity 2-3kg / h; the module start-stop rule is that the compressor, fan, and humidification module run continuously without interruption; the energy consumption control benchmark is ≤110kWh / h. For 10:00-12:00 (medium demand): the core parameter range is compressor frequency 40-45Hz, fan speed 1100-1200r / min, and humidification capacity 2-2.5kg / h; the module start-stop rule is that the compressor and fan run continuously, and the humidification module starts and stops every 30 minutes (adjusted flexibly based on humidity monitoring values); the energy consumption control benchmark is ≤100kWh / h.

[0094] For 13:00-15:00 (high demand): the core parameter range is the same as 8:00-10:00 (compressor frequency 45-50Hz, fan speed 1200-1300r / min, humidification capacity 2-3kg / h); the module start-stop rule is that the compressor, fan, and humidification module run continuously; the energy consumption control benchmark is ≤110kWh / h. For 15:00-17:00 (low demand): the core parameter range is compressor frequency 40-45Hz, fan speed 1000-1100r / min, humidification capacity 1-2kg / h; the module start-stop rule is that the compressor and fan run continuously, and the humidification module starts and stops as needed (starts when the workshop humidity is <45%, and stops when it reaches 50%); the energy consumption control benchmark is ≤90kWh / h. Through the above steps, a complete time-segmented energy-saving control strategy covering the entire daily production cycle is formed, which not only ensures that the process requirements of each time period are met, but also optimizes energy consumption according to equipment characteristics and demand intensity, thereby achieving customized energy-saving goals.

[0095] In one implementation, such as Figure 2 As shown, this application also provides a customized energy-saving air conditioning control device for industrial process needs, comprising:

[0096] The acquisition module 201 is used to acquire real-time process parameters, environmental data and historical energy consumption records of air conditioning operation in industrial production. It uses Kalman filtering algorithm to remove electromagnetic interference and equipment vibration noise in the industrial environment and generates a standardized process-environment-energy consumption related data sequence.

[0097] Processing module 202 is used to convert standardized process-environment-energy consumption correlation data sequences into three-dimensional process load distribution maps, and then process them using fluid dynamics simulation software. An adaptive mesh generation algorithm is used to subdivide high-load process areas, and individualized air conditioning dynamic load prediction models are constructed by combining the differences in thermal and humidity characteristics of different industrial scenarios. Based on the individualized air conditioning dynamic load prediction models, priority factors for different industrial processes are extracted to construct dynamic parameter adjustment matrices. Parallel calculations are performed on the operating states of multiple air conditioning modules to obtain the air conditioning operating parameter combinations under the target energy consumption. Energy consumption peaks, load fluctuation rates, and process-energy consumption matching differences are extracted according to industrial process scenarios to classify energy consumption exceedance risk levels, and a system is constructed that includes process demand intensity, environmental interference coefficient, and equipment energy efficiency degradation. The system generates a multidimensional feature matrix of information; compares real-time energy consumption with historical data from the same process, identifies abnormal signals such as short-term energy consumption spikes and load-energy consumption imbalances, and generates dynamic correction factors for energy consumption optimization using the slope of the energy consumption-process parameter correlation curve; groups users based on the industrial enterprise's production scale, process complexity, and equipment configuration, and uses a gradient boosting tree algorithm to screen key influencing factors of the multidimensional feature matrix and dynamic correction factors for energy consumption optimization; integrates energy consumption optimization parameters, process constraints, and equipment operating limit information to construct a personalized energy-saving control model; based on the target operating parameter output of the personalized energy-saving control model, and combined with the time correlation information of industrial process timing and air conditioning equipment operating characteristics, generates real-time dynamic adjustment instructions and time-segmented energy-saving control strategies.

[0098] The computer-readable storage medium provided in the above embodiments of this application and the customized energy-saving air conditioning control method for industrial process needs provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the application programs stored therein.

[0099] The various embodiments in this application are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for evaluating customized energy-saving air conditioning control methods, electronic devices, electronic equipment, and readable storage media for industrial process needs are basically similar to the embodiments of the customized energy-saving air conditioning control method for industrial process needs described above, and therefore the descriptions are relatively simple. Relevant parts can be referred to in the descriptions of the embodiments of the customized energy-saving air conditioning control method for industrial process needs described above.

Claims

1. A customized energy-saving air conditioning control method for industrial process needs, characterized in that, include: The system acquires real-time process parameters, environmental data, and historical energy consumption records of air conditioning operation in industrial production. It then uses a Kalman filter algorithm to remove electromagnetic interference and equipment vibration noise from the industrial environment, generating a standardized process-environment-energy consumption correlation data sequence. After converting the standardized process-environment-energy consumption correlation data sequence into a three-dimensional process load distribution map, it is processed by fluid dynamics simulation software. An adaptive mesh generation algorithm is selected to subdivide the high-load process area. Combined with the differences in thermal and humidity characteristics of different industrial scenarios, an individualized air conditioning dynamic load prediction model is constructed. Based on the individualized air conditioning dynamic load prediction model, priority factors of different industrial processes are extracted to construct a dynamic parameter adjustment matrix. Parallel calculations are performed on the operating status of multiple air conditioning modules to obtain the combination of air conditioning operating parameters under the target energy consumption. Based on industrial process scenarios, peak energy consumption, load fluctuation rate, and process-energy consumption matching difference values ​​are extracted to classify the risk level of energy consumption exceeding standards, and a multi-dimensional feature matrix containing information on process demand intensity, environmental interference coefficient, and equipment energy efficiency degradation is constructed. By comparing real-time energy consumption with historical data of the same process, abnormal signals such as short-term energy consumption surges and load-energy consumption mismatch are identified, and dynamic correction factors for energy consumption optimization are generated using the slope of the energy consumption-process parameter correlation curve. Users are grouped based on the production scale, process complexity, and equipment configuration of industrial enterprises. The gradient boosting tree algorithm is used to screen key influencing factors of multidimensional feature matrix and dynamic correction factor for energy consumption optimization. Personalized energy-saving control model is constructed by integrating energy consumption optimization parameters, process constraints, and equipment operating limit information. Based on the target operating parameter output of the personalized energy-saving control model, and combined with the time correlation information of industrial process timing and air conditioning equipment operating characteristics, real-time dynamic adjustment instructions and time-sharing energy-saving control strategies are generated.

2. The method as described in claim 1, characterized in that, After converting the standardized process-environment-energy consumption correlation data sequence into a three-dimensional process load distribution map, it is processed by fluid dynamics simulation software. An adaptive mesh generation algorithm is used to subdivide the high-load process area. Combined with the differences in thermal and humidity characteristics of different industrial scenarios, an individualized dynamic air conditioning load prediction model is constructed, including: The standardized process-environment-energy consumption correlation data sequence is converted into a three-dimensional process load distribution map to generate basic load visualization data. The three-dimensional process load distribution map constructs a multi-dimensional correlation visualization map by mapping process parameters to the X-axis, environmental data to the Y-axis, and energy consumption data to the Z-axis. The three-dimensional process load distribution map is imported into the fluid dynamics simulation software for simulation calculation to generate thermal and moisture load field simulation data. The fluid dynamics simulation software simulates the air flow trajectory and heat and moisture exchange efficiency in the industrial space, and quantifies the thermal and moisture load intensity of different process areas. An adaptive mesh generation algorithm is used to subdivide the high-load process area in the thermal and humidity load field simulation data to generate differentiated mesh data. The adaptive mesh generation algorithm identifies the thermal and humidity load density in real time. When the load density exceeds a preset threshold, it automatically triggers mesh refinement. High-density meshes are used for high-load areas and low-density meshes are used for low-load areas. Collect parameters showing differences in thermal and humidity characteristics in different industrial scenarios, construct a scenario-based thermal and humidity characteristic constraint library, and generate scenario constraint data. The scenario-based thermal and humidity characteristic constraint library customizes parameters to meet the unique thermal and humidity requirements of different scenarios. By integrating differentiated grid data with scenario-constrained data, an individualized dynamic load prediction model for air conditioning is constructed.

3. The method as described in claim 1, characterized in that, Based on an individualized dynamic load prediction model for air conditioning, priority factors for different industrial processes are extracted to construct a dynamic parameter adjustment matrix. Parallel calculations are then performed on the operating states of multiple air conditioning modules to obtain the combination of air conditioning operating parameters under the target energy consumption, including: Based on the individualized air conditioning dynamic load prediction model, priority factors of different industrial processes are extracted, and a dynamic parameter adjustment matrix is ​​constructed. The priority factor extraction takes the production criticality, heat and humidity demand urgency, and energy consumption sensitivity of the industrial process as the core dimensions. The dynamic parameter adjustment matrix associates and maps the priority factors of each process with the air conditioning adjustment parameters to generate a parameter adjustment logic framework that can be updated in real time. The dynamic parameter adjustment matrix is ​​input into the parallel computing module to synchronously calculate the operating status of multiple air conditioning modules and generate multiple sets of candidate operating parameters. The parallel computing module adopts a distributed computing architecture and performs parameter iterative calculations simultaneously, taking into account the independent operating characteristics and interactive influence relationships of each air conditioning module. Based on the preset target energy consumption threshold, the candidate operating parameter set is screened and optimized to obtain the air conditioning operating parameter combination under the target energy consumption. The screening process involves constructing a dual-objective evaluation function for energy consumption and process satisfaction, introducing parameter sensitivity analysis, and predicting the energy consumption impact of small fluctuations in key adjustment parameters.

4. The method as described in claim 1, characterized in that, Energy consumption peaks, load fluctuation rates, and process-energy consumption matching differences are extracted based on industrial process scenarios to classify energy consumption exceedance risk levels. A multi-dimensional feature matrix is ​​constructed, including information on process demand intensity, environmental interference coefficients, and equipment energy efficiency degradation. Energy consumption peak, load fluctuation rate, and process-energy consumption matching difference value are extracted according to industrial process scenarios to generate energy consumption feature data. Among them, the energy consumption peak is extracted as the maximum value of air conditioning energy consumption per unit time under the corresponding process scenario, the load fluctuation rate is calculated as the amplitude and frequency of energy consumption change per unit time, and the process-energy consumption matching difference value is obtained by comparing the deviation value between actual energy consumption and theoretical energy consumption required by the process. Based on the numerical range of peak energy consumption, load fluctuation rate, and process-energy consumption matching difference, the risk level of energy consumption exceeding the standard is divided, and risk level identification data is generated. Among them, multiple sets of energy consumption feature threshold ranges are preset, and the extracted energy consumption feature data is compared with the threshold ranges to automatically match the corresponding risk level. Information on process demand intensity, environmental interference coefficient, and equipment energy efficiency degradation in industrial process scenarios is collected. A multi-dimensional feature matrix containing information on process demand intensity, environmental interference coefficient, and equipment energy efficiency degradation is constructed. The multi-dimensional feature matrix uses row vectors to represent different industrial process scenarios, and column vectors to correspond to the feature dimensions of process demand intensity, environmental interference coefficient, and equipment energy efficiency degradation information, respectively. Each matrix element is a specific quantitative value of the feature.

5. The method as described in claim 1, characterized in that, By comparing real-time energy consumption with historical data from the same period and process, abnormal signals such as short-term energy consumption spikes and load-energy consumption imbalances are identified. Dynamic correction factors for energy consumption optimization are generated using the slope of the energy consumption-process parameter correlation curve, including: Retrieve real-time energy consumption data of industrial processes and historical energy consumption data of the same process in the same period to establish data comparison dimensions. Among them, real-time energy consumption data is the instantaneous energy consumption and average energy consumption per unit time of air conditioning in the current process operation cycle, and historical energy consumption data of the same process in the same period is the average energy consumption and fluctuation range under the same season and the same production load in the past. Based on preset energy consumption difference thresholds and load-energy consumption matching benchmarks, real-time energy consumption is compared with historical data of the same process in the same period to identify abnormal signals such as short-term energy consumption surges and load-energy consumption imbalances. Short-term energy consumption surges are determined by monitoring whether the energy consumption increase per unit time exceeds the threshold, and load-energy consumption imbalances are determined by comparing whether the correlation between actual energy consumption changes and process load changes meets the benchmark. Collect energy consumption data and process parameters for the corresponding industrial process, construct an energy consumption-process parameter correlation curve, calculate the slope of the curve at the current process parameter node, and use this slope to generate a dynamic correction factor for energy consumption optimization.

6. The method as described in claim 5, characterized in that, Based on the target operating parameter output of the personalized energy-saving control model, and combined with the time correlation information between the industrial process sequence and the operating characteristics of the air conditioning equipment, real-time dynamic adjustment commands and time-sharing energy-saving control strategies are generated, including: Extract the target operating parameters output by the personalized energy-saving control model, and simultaneously collect the industrial process timing and air conditioning equipment operating characteristics to establish a time-related mapping relationship among the three, generating parameter-time-equipment related data. Based on parameter-time-equipment correlation data, and targeting the real-time operating status of industrial processes, the deviation between the target operating parameters and the actual operating parameters is calculated in real time. Combined with the real-time operating characteristics of the air conditioning equipment, real-time dynamic adjustment instructions are generated. The real-time dynamic adjustment instructions include the parameter adjustment direction, adjustment range, and execution time, and the frequency of instruction generation is synchronized with the frequency of process status changes. Based on the time period division in the industrial process sequence arrangement, and combined with the process demand intensity corresponding to each time period and the differences in the operating characteristics of air conditioning equipment in different time periods, the target operating parameters of each time period are adapted and adjusted in a time-specific manner to generate a time-segmented energy-saving control strategy. The time-segmented energy-saving control strategy clarifies the range of core air conditioning operating parameters, module start-stop rules and energy consumption control benchmarks for each time period.

7. A customized energy-saving air conditioning control device for industrial process needs, characterized in that, The device includes: The acquisition module is used to acquire real-time process parameters, environmental data and historical energy consumption records of air conditioning operation in industrial production. It uses Kalman filtering algorithm to remove electromagnetic interference and equipment vibration noise in the industrial environment and generate a standardized process-environment-energy consumption correlation data sequence. The processing module converts standardized process-environment-energy consumption correlation data sequences into three-dimensional process load distribution maps. After processing with fluid dynamics simulation software, an adaptive mesh generation algorithm is used to subdivide high-load process areas. Combined with the differences in thermal and humidity characteristics across different industrial scenarios, an individualized dynamic load prediction model for air conditioning is constructed. Based on this model, priority factors for different industrial processes are extracted to construct a dynamic parameter adjustment matrix. Parallel calculations are performed on the operating states of multiple air conditioning modules to obtain the combination of air conditioning operating parameters under the target energy consumption. Energy consumption peaks, load fluctuation rates, and process-energy consumption matching differences are extracted according to industrial process scenarios to classify energy consumption exceedance risk levels. A system is constructed that includes process demand intensity, environmental interference coefficient, and equipment energy efficiency degradation information. The system generates a multidimensional feature matrix of energy consumption data; compares real-time energy consumption with historical data from the same period and process, identifies abnormal signals such as short-term energy consumption spikes and load-energy consumption imbalances, and generates dynamic correction factors for energy consumption optimization using the slope of the energy consumption-process parameter correlation curve; groups users based on the industrial enterprise's production scale, process complexity, and equipment configuration, and uses a gradient boosting tree algorithm to screen key influencing factors of the multidimensional feature matrix and dynamic correction factors for energy consumption optimization; integrates energy consumption optimization parameters, process constraints, and equipment operating limit information to construct a personalized energy-saving control model; based on the target operating parameter output of the personalized energy-saving control model, and combined with the time correlation information of industrial process timing and air conditioning equipment operating characteristics, generates real-time dynamic adjustment instructions and time-segmented energy-saving control strategies.

8. An electronic device, characterized in that, include: First processor; and memory for storing executable instructions of the first processor; The first processor is configured to execute the customized energy-saving air conditioning control method for industrial process needs as described in any one of claims 1 to 6 by executing the executable instructions.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the second processor, it implements the customized energy-saving air conditioning control method for industrial process needs as described in any one of claims 1 to 6.

Images