An intelligent optimization control method for industrial air conditioners and waste heat recovery and related equipment

Abstract

The application provides an intelligent optimization control method for industrial air conditioners and waste heat recovery and related equipment, and applies to the technical field of data processing. Through the fusion of industrial air conditioners and waste heat recovery system data by Internet of Things and edge computing, the application generates a standardized parameter sequence through wavelet denoising, converts it into an energy flow network atlas, and subdivides high-energy-consumption working conditions by fuzzy clustering to construct a dynamic energy efficiency coupling model. Based on the model, an optimization equation is constructed by using an improved particle swarm algorithm, and the energy consumption of key nodes is obtained by combining an energy cascade utilization matrix. Then, an energy-saving potential coefficient is generated through a multi-dimensional feature matrix, combined with a production plan grouping operation mode, and a self-adaptive optimization control model is constructed by using deep reinforcement learning to finally generate real-time regulation and control instructions and energy efficiency schemes, so that system collaborative optimization is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to an intelligent optimization control method for industrial air conditioners and waste heat recovery and related equipment. BACKGROUND

[0002] In the field of industrial production, industrial air conditioners are one of the core equipment to ensure the stability of the production environment, and their energy consumption usually accounts for 15%-30% of the total industrial energy consumption. However, a large amount of waste heat (such as heat dissipation of production equipment, heat release of process reaction, etc.) is generated during industrial production. If these waste heat is not effectively recovered, not only energy is wasted, but also environmental heat pollution may be aggravated. In the current industrial scene, the industrial air conditioner and the waste heat recovery system are in independent operation state, lacking a cooperative control mechanism, resulting in significant contradictions and deficiencies in energy utilization between the two systems. The specific problems are as follows:

[0003] The existing technology relies on a single sensor or traditional monitoring method for collecting industrial air conditioner operating parameters (such as return air temperature, compressor power, and refrigerating capacity) and waste heat recovery system data (such as waste heat medium flow, heat exchange efficiency, and medium temperature), without using multi-source data fusion technology, and without effectively filtering equipment vibration and other interference noise, resulting in errors in the original data (such as compressor power signal fluctuating by ±1.5kW due to vibration interference), which cannot provide accurate data support for subsequent energy efficiency analysis and control.

[0004] Industrial production conditions are complex and dynamically changing (such as production load fluctuations and environmental temperature changes), and the existing technology is difficult to accurately subdivide high-energy-consumption conditions, and mostly uses single mechanism modeling or data-driven modeling methods, without combining the advantages of both to build a coupled model, resulting in the model being unable to accurately reflect the energy coupling relationship of the air conditioner-waste heat system, such as being unable to quantify the impact of waste heat quality (temperature gradient, flow stability) on air conditioner energy consumption under different conditions, and thus being difficult to achieve dynamic optimization of system energy efficiency.

[0005] On the one hand, the waste heat recovery system does not utilize waste heat in stages according to its quality, such as treating high-temperature waste heat (80-100℃) and low-temperature waste heat (30-50℃) equally, resulting in insufficient utilization of high-grade waste heat; on the other hand, the air conditioner system and the waste heat recovery system lack coordinated regulation and control, such as when air conditioner load increases suddenly, the waste heat recovery system does not adjust the medium flow in time to improve waste heat supply, resulting in the air conditioner consuming a large amount of electric energy, and the energy loss at key nodes (such as pipeline transmission and heat exchanger) is not accurately identified (such as the fluctuation error of the pipeline heat loss coefficient being more than 0.005kW / m・h), further reducing the overall energy efficiency of the system.

[0006] It should be noted that the information disclosed in the above background section is only used to strengthen the understanding of the background of the present disclosure, and therefore can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0007] Other features and advantages of the present application will become apparent from the following detailed description, taken in conjunction with the accompanying drawings, which illustrate, by way of example, the principles of the application.

[0008] According to an aspect of the present application, an intelligent optimization control method of industrial air conditioner and waste heat recovery is provided, comprising: obtaining industrial air conditioner operating parameters and waste heat recovery system data, based on multi-source data fusion technology of Internet of Things sensors and edge computing gateways, using wavelet denoising algorithm to eliminate equipment vibration interference noise, and generating standardized operating parameter sequence; after converting the standardized operating parameter sequence into a system energy flow network graph, processing it through mechanism-data hybrid modeling software, selecting fuzzy clustering algorithm to subdivide high energy consumption conditions, and constructing a dynamic energy efficiency coupling model; based on the dynamic energy efficiency coupling model, setting the system energy consumption as a condition-time dependent function, using an improved particle swarm algorithm to couple a load prediction correction term to construct an optimization objective equation, extracting conversion coefficients of different waste heat qualities to construct an energy cascade utilization matrix, and performing collaborative calculation on air conditioner-waste heat system energy consumption distribution to obtain key node energy loss value; combining the key node energy loss value, extracting energy consumption peak, waste heat utilization rate, and load matching degree according to production condition scene, dividing energy efficiency level, and constructing a multi-dimensional feature matrix containing operating parameters, load characteristics, and environmental condition information; based on the multi-dimensional feature matrix, comparing real-time energy efficiency with historical data to identify abnormal signals such as sudden increase in energy consumption and insufficient waste heat utilization, and using the slope of the energy efficiency-load curve to generate energy saving potential coefficient; combining the energy saving potential coefficient with production plan to group operating modes, using deep reinforcement learning algorithm to screen key control factors, fusing energy efficiency parameters, equipment status, and production demand information, and constructing an adaptive optimization control model; based on the control strategy output of the adaptive optimization control model, combining dynamic correlation information of air conditioner load and waste heat supply, generating real-time control instructions and control information of energy efficiency improvement scheme.

[0009] In another aspect of the present application, an intelligent optimization control device for industrial air conditioners in cooperation with waste heat recovery includes: a collection module for obtaining industrial air conditioner operating parameters and waste heat recovery system data, based on multi-source data fusion technology of Internet of Things sensors and edge computing gateways, using a wavelet denoising algorithm to eliminate device vibration interference noise, and generating a standardized operating parameter sequence; a processing module for converting the standardized operating parameter sequence into a system energy flow network graph, processing it through a mechanism-data hybrid modeling software, selecting a fuzzy clustering algorithm for high energy consumption condition subdivision, and constructing a dynamic energy efficiency coupling model; based on the dynamic energy efficiency coupling model, setting the system energy consumption as a working condition-time dependent function, using an improved particle swarm algorithm to couple a load prediction correction term to construct an optimization objective equation, extracting conversion coefficients of different waste heat qualities to construct an energy cascade utilization matrix, and performing collaborative calculation on air conditioner-waste heat system energy consumption distribution to obtain key node energy loss values; combining the key node energy loss values, extracting energy consumption peaks, waste heat utilization rates, and load matching degrees according to production working condition scenarios, dividing energy efficiency levels, and constructing a multi-dimensional feature matrix containing operating parameters, load characteristics, and environmental condition information; based on the multi-dimensional feature matrix, comparing real-time energy efficiency with historical data to identify abnormal signals such as sudden energy consumption increase and insufficient waste heat utilization, and using the slope of the energy efficiency-load curve to generate an energy saving potential coefficient; combining the energy saving potential coefficient with production plans to group operating modes, using a deep reinforcement learning algorithm to select key control factors, fusing energy efficiency parameters, device states, and production demand information, and constructing an adaptive optimization control model; based on the control strategy output of the adaptive optimization control model, combining dynamic correlation information of air conditioner load and waste heat supply, generating real-time control instructions and control information of energy efficiency improvement schemes.

[0010] According to still another aspect of the present 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-mentioned intelligent optimization control method for industrial air conditioners in cooperation with waste heat recovery by executing the executable instructions.

[0011] According to still another aspect of the present application, a computer readable storage medium having a computer program stored thereon is provided, the computer program being executed by a second processor to implement the above-mentioned intelligent optimization control method for industrial air conditioners in cooperation with waste heat recovery.

[0012] The industrial air conditioner and waste heat recovery collaborative intelligent optimization control method and related equipment provided by the application take industrial air conditioner and waste heat recovery collaborative optimization as the core, generate a standardized parameter sequence through wavelet denoising by fusing multi-source data through Internet of Things and edge computing; after being converted into an energy flow network atlas, high-energy-consumption conditions are subdivided using fuzzy clustering, and a dynamic energy efficiency coupling model driven by mechanism-data hybrid is constructed. Based on the model, an optimization equation is constructed using an improved particle swarm algorithm, and key node energy consumption is obtained in combination with an energy cascade utilization matrix; then, an energy saving potential coefficient is generated through a multi-dimensional feature matrix, a production plan is combined with a group operation mode, a regulation factor is screened using deep reinforcement learning (Actor-Critic architecture), an adaptive optimization control model is constructed, and finally real-time regulation instructions and energy efficiency schemes are generated, forming a "data acquisition-modeling-optimization-control" closed loop.

[0013] It should be understood that the foregoing general description and the following detailed description are only exemplary and explanatory, and are not limiting to the present disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 A flowchart of an industrial air conditioner and waste heat recovery collaborative intelligent optimization control method provided by an embodiment of the application is shown;

[0015] Figure 2 A structural schematic diagram of an industrial air conditioner and waste heat recovery collaborative intelligent optimization control device provided by an embodiment of the application is shown. DETAILED DESCRIPTION

[0016] The preferred embodiments of the application are described below in conjunction with the accompanying drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the application, and are not used to limit the application.

[0017] The industrial air conditioner and waste heat recovery collaborative intelligent optimization control method according to the exemplary embodiments of the application is described below in conjunction with Figure 1 It should be noted that the following application scenarios are only shown for the purpose of facilitating understanding of the spirit and principles of the application, and the embodiments of the application are not limited in this respect. On the contrary, the embodiments of the application are applicable to any applicable scenarios.

[0018] In an embodiment, Figure 1 A flowchart of an industrial air conditioner and waste heat recovery collaborative intelligent optimization control method according to an embodiment of the application is shown schematically, which includes:

[0019] S101, obtain industrial air conditioner operation parameters and waste heat recovery system data, based on multi-source data fusion technology of Internet of Things sensors and edge computing gateways, use a wavelet denoising algorithm to eliminate device vibration interference noise, and generate a standardized operation parameter sequence.

[0020] In one embodiment, when obtaining industrial air conditioning operating parameters and waste heat recovery system data, the core operating indicators of the two types of systems need to be fully covered. Among them, in addition to the return air temperature (such as 25℃±0.5℃), the refrigeration compressor power (such as 15kW), and the fan speed (such as 1450r / min), the industrial air conditioning operating parameters also include the supply air temperature (such as 18℃±0.3℃), the condenser inlet and outlet water temperature (such as 30℃ / 35℃), the refrigeration capacity (such as 50kW), etc., which directly reflect the refrigeration efficiency and operating state of the air conditioner; in addition to the waste heat medium (such as hot water) flow (such as 8m³ / h), the medium inlet temperature (such as 60℃), and the heat exchanger heat exchange efficiency (such as 85%), the waste heat recovery system data also covers the medium outlet temperature (such as 45℃), the waste heat recovery amount (such as 12kW), and the pipeline pressure loss (such as 0.1MPa), etc., which can accurately reflect the efficiency and system loss of waste heat recovery.

[0021] Based on the multi-source data fusion technology of Internet of Things sensors and edge computing gateways, a sensing and processing network covering the whole system is constructed. Specifically, temperature sensors (precision ±0.1℃) are installed in the return air duct and the supply air duct of the air conditioning unit, power sensors (sampling frequency 1kHz) are installed at the output end of the compressor, and speed sensors are installed at the fan motor; flow sensors (range 0-20m³ / h) and temperature sensors are installed at the inlet and outlet pipes of the heat exchanger of the waste heat recovery system respectively, and an efficiency monitoring module is installed on the heat exchanger shell. These Internet of Things sensors transmit the collected dispersed data to the edge computing gateway in real time. The gateway first performs format verification (such as eliminating abnormal values of temperature sensors exceeding the range of-50℃~100℃), and then converts different types of parameters (temperature, flow, power, etc.) into unified data frames in JSON format, and realizes the spatio-temporal synchronization of air conditioning operating parameters and waste heat recovery data through timestamp alignment (error ≤10ms), ensuring the correlation of multi-source data in time and space dimensions.

[0022] When using wavelet denoising algorithm to eliminate equipment vibration interference noise, different parameters of noise characteristics are accurately processed. Taking the compressor power signal as an example, the original signal is mixed with high-frequency noise generated by the mechanical vibration of the compressor (such as piston reciprocating motion, motor running). Through db4 wavelet basis, the signal is decomposed into 8 scales, and the 7th-8th scale corresponds to the high-frequency noise component above 1000Hz, which presents irregular violent fluctuations (instantaneous amplitude up to ±1.2kW). After eliminating this part of noise, the signal is reconstructed by wavelet inverse transform, so that the fluctuation range of power data is narrowed from the original ±1.5kW to within ±0.3kW, effectively retaining the trend characteristics of power signal changes with load (such as when the production load increases, the power increases from 15kW to 18kW smoothly), avoiding the interference of noise on subsequent energy efficiency analysis.

[0023] The data is normalized and structured when generating the standardized operation parameter sequence. First, the processed data is divided into time series at 1-minute intervals, and each time node corresponds to a complete set of parameters (e.g., at 10:00:00, the return air temperature is 25.2°C, the compressor power is 15.1kW, and the waste heat medium flow is 7.9m³ / h, etc.). Second, the parameter units are unified, such as temperature in °C, power in kW, and flow in m³ / h, to avoid calculation errors caused by unit confusion. Finally, the min-max normalization method is used to map each parameter value to the 0-1 interval (e.g., the compressor power range is 10-20kW, and 15kW corresponds to 0.5), forming a three-dimensional sequence data containing "timestamp (2024-05-2010:00:00), parameter identifier ('air conditioner_compressor power'), standardized value (0.5)". This sequence can be directly imported into a mechanism-data hybrid modeling software, providing a high-quality data foundation for subsequent construction of system energy flow network atlas.

[0024] In S102, after converting the standardized operation parameter sequence into a system energy flow network atlas, the mechanism-data hybrid modeling software is used to process and select a fuzzy clustering algorithm to subdivide high-energy consumption conditions, and to construct a dynamic energy efficiency coupling model.

[0025] In one embodiment, the air conditioner energy consumption data, waste heat medium flow data, and heat exchange efficiency parameters in the standardized operation parameter sequence are extracted and converted to generate energy flow node association groups, network transmission loss coefficients, and parameter space-time matching degrees. The hourly electricity consumption of the air conditioner compressor (e.g., 120kWh), the hourly circulation flow of the waste heat medium (hot water) (e.g., 480m³), and the real-time heat exchange efficiency of the heat exchanger (e.g., 82%) are extracted from the standardized sequence. The air conditioning unit, waste heat exchanger, and pipeline are mapped to energy flow nodes through a directed graph model to generate node association groups (e.g., the energy transfer link of "compressor → condenser → waste heat exchanger → user end"). The network transmission loss coefficient is calculated based on the pipe material and length (e.g., the heat loss coefficient of steel pipe is set to 0.03kW / m・h). The parameter space-time matching degree is obtained by time series alignment and spatial position matching (e.g., the time matching degree of air conditioner energy consumption peak and waste heat generation peak is 85%, and the spatial transmission distance matching degree is 90%).

[0026] The energy flow node association group, network transmission loss coefficient, and parameter space-time matching degree are processed to generate mechanism modeling constraint conditions, data fitting error threshold, and clustering feature extraction dimension. According to the energy conservation law in the node association group, the mechanism modeling constraint conditions are set (such as compressor output energy = refrigeration capacity + waste heat emission + transmission loss, and the error allowed range is ≤5%); based on the fluctuation range of the network transmission loss coefficient, the data fitting error threshold is set (such as the fitting error of heat loss coefficient ≤0.002kW / m・h); combined with the key influencing factors of parameter space-time matching degree, the clustering feature extraction dimension is determined as 3 dimensions (time synchronization dimension, space transmission dimension, and energy loss dimension).

[0027] Based on the mechanism modeling constraint conditions, data fitting error threshold, and clustering feature extraction dimension, the high energy consumption path, waste heat utilization bottleneck, and working condition mutation signal in the system energy flow network graph are classified and subdivided to generate a high energy consumption working condition feature library. In the system energy flow network graph, the high energy consumption path of “compressor high load operation → pipeline transmission loss exceeds the standard → waste heat exchanger efficiency is lower than 70%” is identified, and is subdivided according to the daily running time length ratio (such as the path with a daily appearance time length ≥6h is classified as A, and 3-6h is classified as B); for the waste heat utilization bottleneck (such as insufficient medium flow leading to insufficient heat exchange), it is subdivided according to the flow gap size (such as a gap ≥20% is a serious bottleneck, and 10%-20% is a general bottleneck); for the working condition mutation signal (such as a sudden increase in production load leading to a sudden rise in air conditioning energy consumption of more than 20%), it is subdivided according to the mutation amplitude and duration (such as an amplitude ≥30% and a duration ≥10min is an emergency mutation); the above subdivision results are integrated into a high energy consumption working condition feature library containing 12 types of features, each type of feature including energy flow intensity, duration, associated node state, and other parameters.

[0028] The high energy consumption condition characteristic library is fused and coupled to generate a dynamic energy efficiency coupling model containing energy efficiency correlation equation, condition dynamic conversion rule, and system coordination constraint condition. The model is a mechanism-data hybrid driven multivariate coupling model, which is hierarchically divided into physical layer (corresponding to actual device energy transmission), characteristic layer (corresponding to condition characteristic parameter), and decision layer (corresponding to energy efficiency optimization target). The neural network architecture of the model is “input layer-hidden layer-output layer”, the input layer has 8 nodes (including air conditioner energy consumption, waste heat flow, etc.), the hidden layer has 2 layers (each layer has 16 neurons), and the output layer has 3 nodes (energy efficiency value, conversion probability, and constraint satisfaction degree). The energy efficiency correlation equation is set as E = αP + βQ - γL (where E is the system comprehensive energy efficiency, P is the air conditioner power, Q is the waste heat recovery amount, L is the transmission loss, and α, β, and γ are coefficients obtained through data fitting, which are 0.6, 0.3, and 0.1, respectively). The condition dynamic conversion rule adopts a Markov chain model, and the probability of conversion from a class A high energy consumption condition to a class B condition is set as 0.3 (satisfying a 10% reduction in transmission loss). The system coordination constraint condition is set as “air conditioner refrigeration capacity ≥ production cold load demand”, “waste heat recovery amount ≥ secondary energy demand”, and “total energy consumption ≤ industry benchmark value 150 kWh / ton of product”.

[0029] In S103, based on the dynamic energy efficiency coupling model, the system energy consumption is set as a condition-time dependent function, an improved particle swarm algorithm is used to couple a load prediction correction term to construct an optimization target equation, a conversion coefficient of different waste heat qualities is extracted to construct an energy cascade utilization matrix, and the air conditioner-waste heat system energy consumption distribution is calculated to obtain the key node energy loss value.

[0030] In an embodiment, the condition correlation parameters in the dynamic energy efficiency coupling model, the time series energy consumption data, and the system coordination constraint condition are extracted and adapted to generate an energy consumption function variable group, a time weight coefficient, and a condition conversion threshold. The condition correlation parameters are extracted from the physical layer and the characteristic layer of the dynamic energy efficiency coupling model, including air conditioner refrigeration capacity 40 kW and waste heat generation amount 25 kW corresponding to 80% of production load, air conditioner refrigeration capacity 25 kW and waste heat generation amount 15 kW corresponding to 50% of production load, and air conditioner power compensation value 5 kW corresponding to 35℃ of ambient temperature, etc. These parameters reflect the matching relationship of system energy under different conditions. The time series energy consumption data covers a complete period, including hourly energy consumption from 8:00 to 18:00 on weekdays (peak value 150 kWh at 12:00, secondary peak value 130 kWh at 10:00), valley value 30 kWh at 1:00 in the morning, and weekend energy consumption data (e.g., average energy consumption 60 kWh on Saturday) to reflect energy consumption rules at different times.

[0031] The system coordination constraint conditions include that the waste heat recovery amount is not less than 30% of the total air conditioning energy consumption, the deviation of air conditioning refrigeration capacity and production cold load is less than 5%, and the pipeline transmission loss rate is less than 8%. Through min-max normalization, the parameters are mapped to the 0-1 interval, and the Pearson correlation coefficient is used to screen the variables that are strongly related to energy consumption, to generate the energy consumption function variable group (4 core variables of production load, ambient temperature, air conditioning power and waste heat medium flow); based on the proportion of energy consumption in each period to the total daily energy consumption, the time weight coefficient is calculated, the weight is set to 0.8 in the daytime production period (8:00-18:00) because the energy consumption accounts for 85%, the weight is set to 0.2 in the non-production period at night (18:00-8:00 the next day) because the energy consumption accounts for 15%, and the weight is increased to 0.85 in the production peak period (10:00-14:00); according to the analysis of the critical value of the historical working condition conversion, in addition to the production load being greater than or equal to 70% as high load and less than or equal to 30% as low load, a medium load working condition (30%-70%) is added, and an auxiliary threshold is set in relation to the ambient temperature (such as triggering the intensified heat dissipation mode when the ambient temperature is greater than or equal to 32°C in the high load working condition).

[0032] The energy consumption function variable group, time weight coefficient and working condition conversion threshold are processed to generate the optimization target equation structure, particle swarm algorithm parameter range and load prediction correction rule. Based on the energy consumption function variable group (production load x, ambient temperature y, air conditioning power z and waste heat medium flow w), the optimization target equation structure is constructed in combination with the system coordination target, in addition to minf(x, y, z, w) = αz-βq+γl (α = 0.6, β = 0.3, γ = 0.1), a constraint term h(x, y, w) (such as a penalty coefficient when the waste heat medium flow is insufficient) is introduced, so that the equation is more suitable for actual optimization requirements; the particle swarm algorithm parameter range is dynamically adjusted according to the working condition characteristics, the population size is set to 100 in the high load working condition and 50 in the low load working condition, the inertia weight is set to 0.9 (global search) in the early iteration (1-30 times) and 0.5 (local convergence) in the later iteration (71-100 times), and the learning factors c1 and c2 are adjusted to 2.2 in the high potential optimization interval (energy saving space greater than or equal to 15%) to enhance the learning ability; the load prediction correction rule is set in stages, when the deviation of actual load and predicted load is 5%-10%, the correction coefficient k = 1.1; when the deviation is greater than or equal to 10%, k = 1.2, and a secondary correction is triggered in combination with the working condition conversion threshold (such as an additional compensation coefficient of 0.05 when switching from low load to medium load).

[0033] Based on the optimization target equation structure, particle swarm algorithm parameter range, load prediction correction rule, the temperature gradient, flow stability, and energy conversion efficiency of different waste heat qualities are classified and quantitatively processed to generate an energy cascade utilization matrix. In addition to high (80-100℃), medium (50-80℃), and low (30-50℃) levels of temperature gradient, the high level is further subdivided into 80-90℃ and 90-100℃, and the medium level is further subdivided into 50-65℃ and 65-80℃ to more accurately reflect the waste heat grade. The flow stability is evaluated by the fluctuation coefficient within 30 minutes, and ≤5% is stable (such as waste heat medium flow 8m³ / h±0.4m³ / h), 5%-10% is relatively stable (8m³ / h±0.8m³ / h), and >10% is unstable (8m³ / h±1.0m³ / h), and the fluctuation frequency is recorded (such as unstable state appearing ≥3 times per hour is downgraded). The energy conversion efficiency is subdivided according to the type of heat exchanger, and the plate heat exchanger ≥85% is high efficiency, 75%-85% is medium efficiency, the tube heat exchanger ≥80% is high efficiency, and 65%-80% is medium efficiency to ensure that the quantitative standard matches the equipment characteristics. Based on the optimization target equation, the comprehensive benefit value of each level of waste heat is calculated, the particle swarm algorithm is iteratively optimized, and a 3x3x3 energy cascade utilization matrix is generated, in which the waste heat corresponding to the matrix element value 0.95 is 90-100℃, stable flow, and high efficient conversion of plate heat exchanger, the waste heat corresponding to 0.85 is 80-90℃, relatively stable flow, and high efficient conversion of tube heat exchanger, and the waste heat corresponding to 0.3 is 30-50℃, unstable flow, and low efficient conversion, and the matrix element value directly reflects the priority of waste heat utilization.

[0034] The energy cascade utilization matrix is cooperatively calculated and iteratively optimized to generate key node energy loss values including node instantaneous energy consumption value, cumulative loss energy, and system energy efficiency proportion. The improved particle swarm algorithm is used to perform 100 times of iterative calculation on the matrix. The first 50 times of iteration are based on global optimal solution search, and the last 50 times are combined with load prediction correction rules to dynamically adjust the search direction (for example, in the 60th iteration, due to the load prediction deviation of 12%, the correction coefficient k=1.2 is introduced to adjust the particle velocity); the cooperative calculation covers 12 key nodes (compressor, condenser, waste heat exchanger, air supply pipeline, etc.), and the instantaneous energy consumption value of each node is accurate to the minute level (for example, the instantaneous energy consumption of the compressor at 14:00:00 is 22kW, and at 14:01:00 is 21.8kW), the cumulative loss energy is counted according to the day, week, and month cycle (for example, the pipeline transmission node accumulates 8.5kWh of loss energy per day, and 58kWh per week), and the system energy efficiency proportion is calculated in combination with the node energy input-output ratio (for example, the waste heat exchanger input energy is 50kWh, and the output utilization is 42kWh, and the energy efficiency proportion is 28%). The final output of the key node energy loss value is attached with a loss reason label (for example, the sudden increase of 2kW in the instantaneous energy consumption of the compressor is marked as "load fluctuation", and the cumulative loss of the pipeline exceeds the standard and is marked as "aging of insulation layer"), forming a complete evaluation system including real-time monitoring, historical accumulation, and loss attribution, and providing accurate targeting for subsequent energy efficiency optimization.

[0035] S104, in combination with the key node energy loss value, the energy consumption peak value, the waste heat utilization rate, and the load matching degree are extracted according to the production working condition scene, the energy efficiency level is divided, and a multi-dimensional feature matrix including operation parameters, load characteristics, and environmental condition information is constructed.

[0036] In one embodiment, the instantaneous loss peak value, cumulative loss proportion, system energy efficiency deviation in the key node energy loss value are extracted and quantified, and the energy consumption characteristic parameters, waste heat utilization efficiency value, load matching deviation coefficient are generated. From the 12 key nodes (compressor, condenser, waste heat exchanger, air supply pipeline, etc.), the instantaneous loss peak value is extracted at minute level granularity, including 22kW of compressor at 14:00, 3.5kW of pipeline transmission node at 10:00, 5.2kW of condenser at 16:00, 2.8kW of waste heat exchanger at 12:00, etc., to ensure the capture of the loss peak of each node; the cumulative loss proportion is calculated for different periods (day, week, month), such as the cumulative loss of compressor 220kWh, the total loss of system 489kWh, the proportion 45%, the cumulative loss of waste heat exchanger 88kWh, the proportion 18%, and the weekly cumulative proportion (compressor 42%, waste heat exchanger 19%) is recorded to reflect the trend; the system energy efficiency deviation is calculated by the ratio of actual energy efficiency to benchmark value (such as industry secondary energy efficiency standard 3.2kW・h / kW・h), when the actual energy efficiency is 2.94kW・h / kW・h, the deviation is (2.94-3.2) / 3.2=-0.08, if the actual energy efficiency is 3.36, the deviation is +0.05. The standardized processing adopts min-max method, which maps the instantaneous loss peak value (range 0-25kW) to the interval 0-1 (22kW corresponds to 0.88, 3.5kW corresponds to 0.14), the waste heat utilization efficiency value is calculated according to "waste heat recovery amount / waste heat generation amount" (such as generation amount 120kWh, recovery amount 90kWh, efficiency value 0.75), and the load matching deviation coefficient is calculated according to "(air conditioning load-waste heat supply) / air conditioning load" (load 200kW, supply 176kW, deviation 0.12).

[0037] The energy consumption characteristic parameters, waste heat utilization efficiency values, and load matching deviation coefficients are processed to generate production working condition scene classification labels, energy efficiency level division thresholds, and characteristic parameter normalization rules. In addition to S1 (high load persistence, 8:00-18:00, load 80%-90%), S2 (low load intermittence, 0:00-6:00, load 20%-30%), and S3 (fluctuating load, 6:00-8:00 and 18:00-20:00, load fluctuation range ±15%), S4 (extreme load, such as load exceeding 110% during the summer high temperature period) is added, and the labels correspond respectively; the energy efficiency level division thresholds are subdivided according to the waste heat utilization efficiency values, A level (≥0.8, sufficient recovery), B level (0.6-0.8, basically up to standard), C level (0.4-0.6, insufficient utilization), and D level (<0.4, serious waste), and the reward and punishment rules are associated (for example, A level is rewarded with 20 points for energy efficiency optimization, and D level is deducted by 30 points); the characteristic parameter normalization rules are clear: the load matching deviation coefficient is mapped to the interval of -1 to 1 (positive deviation is waste heat surplus, such as 0.12 representing surplus of 12%; negative deviation is insufficient, such as -0.08 representing insufficient of 8%), the energy consumption characteristic parameters retain 3 decimal places to ensure accuracy, and all rules are adapted to different seasons (summer air conditioning load is high, and the deviation coefficient threshold is relaxed to ±0.15).

[0038] Based on the production working condition scene classification label, energy efficiency level division threshold, feature parameter normalization rule, the running parameter fluctuation range, load dynamic change trend, environmental condition influence weight are classified and associated processed, to generate multi-dimensional feature association dataset. For S1 scene, the air conditioner return air temperature fluctuation range (24-26℃, standard deviation 0.3℃), production load dynamic trend (80%-90%, fluctuation ≤2% per hour), environmental temperature influence weight (0.3, air conditioner energy consumption increases 2% for every 1℃ temperature rise), are extracted, associated with energy efficiency level B (efficiency value 0.72), load matching deviation coefficient 0.05 (surplus 5%). S2 scene extracts return air temperature (22-23℃, standard deviation 0.2℃), load trend (20%-30%, intermittent shutdown 1-2 times / h), environmental humidity influence weight (0.2, heat transfer efficiency decreases by 3% when humidity >60%), associated with A level (efficiency value 0.85), deviation coefficient-0.03 (deficiency 3%). S3 scene extracts return air temperature (23-25℃, standard deviation 0.5℃), load trend (fluctuation amplitude ±15%), environmental wind speed influence weight (0.15, heat dissipation efficiency increases by 5% when wind speed >3m / s), associated with B level (0.78), deviation coefficient 0.08; S4 scene extracts return air temperature (26-28℃), load trend (110%-120%), environmental comprehensive influence weight (0.4), associated with C level (0.55), deviation coefficient-0.12. The dataset contains 100 groups of samples (S1:40 groups, S2:20 groups, S3:30 groups, S4:10 groups), each group has 8 features (2 energy consumption characteristic parameters, 1 efficiency value, 1 deviation coefficient, 2 running parameter ranges, 1 trend, 1 weight).

[0039] The multi-dimensional feature correlation data set is integrated and dimensionally optimized to generate a multi-dimensional feature matrix containing a running parameter dynamic interval, a load characteristic classification vector, and an environmental condition influence coefficient. The running parameter dynamic interval is integrated, including a fan speed of 1200-1450 r / min, a waste heat medium flow of 6-10 m³ / h, and a heat exchanger inlet and outlet temperature difference of 15-25 ℃, except for a compressor power of 15-20 kW. The load characteristic classification vector adopts one-hot encoding, with high load [1, 0, 0, 0], medium load [0, 1, 0, 0], low load [0, 0, 1, 0], and extreme load [0, 0, 0, 1]. The environmental condition influence coefficient is refined into temperature (0.35), humidity (0.25), wind speed (0.1), atmospheric pressure (0.05), and light intensity (0.25, significant in summer). Dimension reduction is performed using principal component analysis, and 5 principal components (cumulative contribution rate 86.7%) with eigenvalues greater than 1 are selected from the 8-dimensional features, of which the first principal component (contribution rate 32.1%) mainly reflects the load intensity, and the second principal component (21.3%) reflects the waste heat utilization efficiency. The final multi-dimensional feature matrix is 100x5, with each row corresponding to a group of scenes. For example, the S1 scene sample is [0.82, 0.65, 0.12, 0.35, 1], and the matrix can be directly input into the subsequent model for energy saving potential analysis.

[0040] S105, based on the multi-dimensional feature matrix, the real-time energy efficiency is compared with the historical data to identify abnormal signals such as sudden increase in energy consumption and insufficient waste heat utilization, and the energy saving potential coefficient is generated using the energy efficiency-load curve slope.

[0041] In an embodiment, the dynamic interval of operating parameters, load characteristic classification vector, and environmental condition influence coefficient in the multi-dimensional feature matrix are extracted and quantitatively processed to generate real-time energy efficiency evaluation value, historical energy efficiency benchmark value, and feature parameter deviation degree, wherein the historical energy efficiency benchmark value is weighted and calibrated according to the matching degree of the same period production working condition and environmental condition. The dynamic interval of operating parameters extracted from the multi-dimensional feature matrix includes, in addition to the air conditioner compressor power 15-20 kW and the waste heat medium flow 6-10 m³ / h, the fan speed 1200-1450 r / min, the heat exchanger inlet and outlet temperature difference 15-25 ℃, etc.; the real-time energy efficiency evaluation value is obtained by comprehensive calculation of multiple parameters, such as 2.78 kW·h / kW·h calculated according to “energy efficiency=refrigerating capacity / input power” combined with the refrigerating capacity 50 kW and the input power 18 kW, and rounded to two decimal places as 2.8. The historical data extraction covers the complete period of the same period in the past three months (such as Monday to Friday 8:00-18:00), the production working condition matching degree is refined to the number of equipment running (such as 3 compressors are started at present and in the history, the matching degree is 0.9), and the production shift (such as both are day shift, the matching degree is 1.0), and the average value is 0.95; the environmental condition matching degree covers temperature (current 28 ℃ and history 27 ℃, matching degree 0.95), humidity (current 60% and history 58%, matching degree 0.98), and the average value is 0.96, and then the production working condition (weight 0.6) and the environmental condition (weight 0.4) are weighted and calibrated to the historical benchmark value (before calibration 3.0, after calibration 3.0×(0.95×0.6+0.96×0.4)=2.86). The feature parameter deviation degree is calculated for each key parameter, such as the deviation degree of real-time waste heat medium flow 7.5 m³ / h and interval average 8 m³ / h is (7.5-8) / 8=-0.06, and the deviation degree of real-time environmental temperature coefficient 0.35 and matrix average 0.33 is 0.06.

[0042] The real-time energy efficiency evaluation value, historical energy efficiency benchmark value, and characteristic parameter deviation degree are processed to generate energy efficiency comparison difference value, abnormal signal identification threshold, and fluctuation trend judgment rule. The abnormal signal identification threshold is set as a dynamic threshold interval according to the production load intensity. The energy efficiency comparison difference value includes not only the direct difference between real-time and history (2.8-2.86=-0.06), but also the relative difference (-0.06 / 2.86≈-0.021) to reflect the deviation proportion. The abnormal signal identification threshold is further divided into more detailed load intervals: super-high load (100%-120%) threshold ±0.35, high load (80%-100%) ±0.3, medium-high load (65%-80%) ±0.25, medium load (50%-65%) ±0.2, low load (30%-50%) ±0.18, super-low load (<30%) ±0.15, and each interval is associated with an environmental temperature correction (e.g., threshold increases by 0.05 in summer high temperature). The fluctuation trend judgment rule adds combination conditions of continuous time length and amplitude, such as energy efficiency decrease at 5 consecutive time points (every 10 minutes), and cumulative amplitude exceeding 0.6, or single time point amplitude exceeding 0.3 with characteristic parameter deviation degree exceeding ±0.1, which are all judged as significant downward trend.

[0043] Based on the energy efficiency comparison difference value, abnormal signal identification threshold, and fluctuation trend judgment rule, the energy consumption sudden increase amplitude, waste heat utilization gap, and energy efficiency-load curve shape are classified and slope calculation processed to generate energy saving potential evaluation dataset. The slope calculation introduces a decay coefficient of equipment operation time for correction. The energy consumption sudden increase amplitude is classified according to duration: instantaneous sudden increase (1-5 minutes, such as 150kWh rising to 200kWh), short-term sudden increase (5-30 minutes), and continuous sudden increase (>30 minutes), and the sudden increase causes (such as production load sudden increase, equipment failure) are marked. The waste heat utilization gap is subdivided according to gap proportion: mild gap (<10%, such as demand 140kW, actual 128kW), moderate gap (10%-20%, such as actual 120kW), and severe gap (>20%, such as actual 100kW), and the gap duration (such as severe gap lasting 40 minutes) is recorded simultaneously. The energy efficiency-load curve shape is classified according to slope range: strong negative correlation (slope < -0.03), weak negative correlation (-0.03 to -0.01), no significant correlation (-0.01 to 0.01), and positive correlation (>0.01). When calculating the slope, the decay coefficient is 1.0 for equipment running within 1 year, 0.95 for 1-3 years, 0.9 for 3-5 years, and 0.85 for more than 5 years (such as a 4-year-old equipment, original slope -0.02, corrected slope -0.018). The dataset includes the above classification results and original calculation values, with a total of 12 characteristic dimensions.

[0044] The energy saving potential evaluation dataset is integrated and quantitatively converted to generate an energy saving potential coefficient including instantaneous energy saving space, cumulative energy saving potential and dynamic adjustment coefficient. The dynamic adjustment coefficient is associated with the real-time load rate of the waste heat recovery equipment for dynamic correction. The instantaneous energy saving space is calculated according to the sudden increase type: the instantaneous sudden increase takes 50% of the difference (200-150=50kWh) between the peak value and the reference value as the space (25kWh), the short-term sudden increase takes 60% of the average difference, and the continuous sudden increase takes 80%. The cumulative energy saving potential is calculated according to the daily, weekly and monthly cycles. For example, when the daily light gap is accumulated by 80kWh, 120kWh and 200kWh, the weighted (weight 0.3, 0.5, 0.8) calculation is (80x0.3+120x0.5+200x0.8)=234kWh, which is converted to the potential value 234 / total energy consumption of the day 1000=0.234. The dynamic adjustment coefficient is first set according to the load rate (load rate < 50% is 0.8, 50%-80% is 1.0, and > 80% is 1.2), and then associated with the equipment health degree (such as health score 90 points corresponding to correction factor 0.95), and finally the coefficient = basic value x health factor (such as load rate 80% 1.0x0.95=0.95). The energy saving potential coefficient is presented in the form of a three-dimensional array (such as [25kWh, 0.234, 0.95]), with calculation basis labels (such as "based on 33% instantaneous sudden increase, severe gap 40kW").

[0045] In S106, the energy saving potential coefficient is combined with the production plan to group the operation modes, the key regulation factors are screened by using deep reinforcement learning algorithm, the energy efficiency parameters, equipment state and production demand information are fused, and the self-adaptive optimization control model is constructed.

[0046] In an embodiment, the key production parameters affecting the system operation mode are extracted in combination with the production plan, and the operation modes are divided into high potential group, medium potential group and low potential group according to the energy efficiency optimization level, wherein each group of modes corresponds to a unique optimization priority label. When the key production parameters are extracted in combination with the production plan, the core influencing factors covering the whole production process are covered to ensure the strong correlation of the parameters with the system operation mode. In addition to the daily production shift (such as three shifts: early shift 8:00-16:00, middle shift 16:00-24:00, night shift 0:00-8:00), product capacity (such as 800 tons / day, including capacity distribution in each period: early shift 300 tons, middle shift 350 tons, night shift 150 tons) and equipment running number (such as 3 air conditioning units, of which 2 are main units and 1 is a standby unit), the production process type (such as mechanical processing type requiring stable cold load and chemical type requiring fluctuating cold load), raw material input (such as 500 tons of raw material per day corresponding to 15% increase of cold load demand) and other parameters are also included.

[0047] When dividing the energy efficiency optimization level, the dynamic correlation of seasonal characteristics and production load is combined to refine the group determination standard. The high potential group (energy saving potential coefficient ≥ 0.3) includes production peak period (such as early shift 10:00-14:00, mid-shift 18:00-22:00) running mode in addition to summer high temperature period (environmental temperature ≥ 30℃) running mode. In this scenario, the air conditioning cold load demand is more than 180kW, and the waste heat generation increases synchronously, and the energy saving space is significant. The corresponding priority label is "P1", and the optimization priority is to allocate calculation resources and control authority. The medium potential group (0.15-0.3) includes production flat peak period (such as early shift 8:00-10:00, mid-shift 16:00-18:00) running mode in addition to spring and autumn seasons regular mode (environmental temperature 15-25℃). The cold load demand is 120-180kW, and the waste heat utilization rate is maintained at 70%-80%. The corresponding priority label is "P2", and the optimization strategy focuses on the balance between stability and energy saving. The low potential group (<0.15) includes production low valley period (such as night shift 0:00-6:00) and equipment maintenance period running mode in addition to winter low load mode (environmental temperature <15℃). The cold load demand is <120kW, and the waste heat utilization rate is more than 80%. The energy saving space is limited, and the corresponding priority label is "P3". The optimization mainly focuses on ensuring the basic operation demand.

[0048] The energy saving potential coefficient, energy efficiency parameters and production demand information of different potential groups are compared and analyzed to generate characteristic difference data between groups. When comparing and analyzing the core parameters of different potential groups, three dimensions of data distribution, influencing factors and optimization direction are expanded to ensure accurate quantification of characteristic differences. The core parameters of the high potential group (P1) include energy saving potential coefficient mean value 0.35, key node energy loss value mean value 120kWh (compressor node loss ratio 45%, pipeline transmission node ratio 25%), waste heat utilization rate 65%, production cold load 200kW, in addition to equipment running time (16 hours per day), and environmental temperature mean value 32℃. The corresponding parameters of the medium potential group (P2) are energy saving potential coefficient mean value 0.22, key node energy loss value mean value 80kWh (compressor ratio 40%, pipeline ratio 20%), waste heat utilization rate 75%, production cold load 150kW, equipment daily running 12 hours, and environmental temperature mean value 22℃. The corresponding parameters of the low potential group (P3) are energy saving potential coefficient mean value 0.1, key node energy loss value mean value 50kWh (compressor ratio 35%, pipeline ratio 15%), waste heat utilization rate 85%, production cold load 80kW, equipment daily running 8 hours, and environmental temperature mean value 10℃.

[0049] Further refine the correlation analysis of the difference between groups, and determine the difference causes and optimization breakthrough. The high-potential group is 59% higher than the medium-potential group in energy-saving space (0.35-0.22) / 0.22≈59%, mainly due to high ambient temperature leading to high load operation of air conditioner compressor, resulting in a sharp increase in energy consumption; the low utilization rate of waste heat is 13% (75%-65%), due to the amount of waste heat generated during high temperature period exceeding the processing capacity of the recovery equipment, resulting in waste of part of the waste heat; the production cold load is 33% higher (200-150) / 150≈33%, affected by the superposition of product capacity peak and process heat dissipation demand. The medium-potential group is 120% higher than the low-potential group in energy-saving space (0.22-0.1) / 0.1=120%, because the production load and ambient temperature are moderate, there is still some energy consumption optimization space; the low utilization rate of waste heat is 11.8% (85%-75%) / 75%≈11.8%, affected by frequent start-stop of equipment; the production cold load is 87.5% higher (150-80) / 80=87.5%, related to capacity expansion and process continuous operation.

[0050] Integrate and process the characteristic difference data, energy-saving potential coefficient, energy efficiency parameters, equipment state and production demand information to generate a multi-dimensional model training data set, wherein the energy efficiency parameters include key node energy loss value and waste heat utilization rate. When integrating data, it is necessary to ensure sample coverage and dimension integrity to provide comprehensive support for model training. The number of samples is allocated according to the optimization demand of potential group and the availability of data. The high-potential group collects 400 samples (including summer temperature intervals and production peak period data) due to complex scene and high optimization demand; the medium-potential group collects 300 samples (including spring and autumn different weather, production flat peak period data); the low-potential group collects 300 samples (including winter different low temperature interval, production low valley and maintenance period data).

[0051] The sample dimensions should cover all factors that affect system control. In addition to the energy saving potential coefficient, the energy loss values of 12 key nodes (compressor, condenser, waste heat exchanger, etc.), waste heat utilization rate, equipment health degree (scored on a 10-point scale, e.g. 9 points for good, 6 points for maintenance), and production cold load, environmental parameters (temperature, humidity, wind speed), equipment operating parameters (compressor frequency, fan speed, waste heat medium flow), and production process parameters (raw material input rate, product output rate), a total of 20 dimensions are included. In the data preprocessing stage, the energy loss value (0-200 kWh), production cold load (50-250 kW), and other continuous parameters are mapped to the 0-1 interval using min-max normalization; discrete parameters such as equipment health degree (1-10 points) and production shift (1-3 representing early, mid, and late shifts) are one-hot encoded; and parameters such as environmental temperature (-5-40°C) that are greatly affected by external factors are standardized using Z-score. Finally, a multi-dimensional model training dataset of 1000 valid samples is formed, and the training set (70%), validation set (20%), and test set (10%) are used for subsequent model construction.

[0052] Based on the multi-dimensional model training dataset, a deep reinforcement learning algorithm is used to select key control factors that have a significant impact on control effectiveness, and a preliminary self-adaptive optimization control model containing a policy network and a value network is constructed. The model selection uses the Actor-Critic architecture in deep reinforcement learning, which adapts to the system's dynamic optimization needs and is designed at multiple levels to match data processing and decision logic. The state input layer is responsible for receiving 20 feature dimensions of the multi-dimensional dataset and converting the standardized parameters into a tensor format that the model can recognize; the feature extraction layer uses a fully connected neural network to mine deep relationships between features, capturing the implicit mapping relationship between the energy saving potential coefficient and the control factors; the decision output layer is divided into policy output (Actor) and value evaluation (Critic), which respectively implement control instruction generation and control effectiveness evaluation.

[0053] In terms of model structure, both the policy network (Actor) and the value network (Critic) use a 3-layer fully connected neural network. The input layer has 20 neurons corresponding to the 20 feature dimensions; the first hidden layer has 40 neurons with a ReLU activation function to enhance non-linear fitting ability, and the second hidden layer has 20 neurons with the same ReLU activation function; the output layer of the policy network has 8 neurons corresponding to the 8 key control factors (compressor frequency, fan speed, waste heat medium flow, heat exchanger start-stop state, air supply temperature, condenser water temperature, valve opening, and standby equipment start-stop), with a Softmax activation function to output the action probability of each control factor; the value network has 1 neuron with a linear activation function to output the value evaluation of the current state.

[0054] The model-related parameters are combined with system characteristics and training requirements. The initial learning rate is set to 0.001 to ensure rapid parameter updates at the beginning of training. The discount factor is set to 0.95 to emphasize long-term optimization benefits. The experience replay pool capacity is set to 10,000 to store historical training samples to break data correlation. The batch sampling size is set to 32 to balance training efficiency and stability. The target network update frequency is set to update every 100 steps to avoid training shocks. Key control factors are selected through feature importance analysis and gradient descent method. Finally, the compressor frequency (contribution 35%, directly affecting the core of air conditioning energy consumption), waste heat medium flow (contribution 25%, determining the waste heat recovery efficiency), heat exchanger start-stop state (contribution 20%, controlling the waste heat utilization path), and air supply temperature (contribution 10%, affecting the cold load matching degree) are determined as the core control factors, and the remaining factors are used as auxiliary control items.

[0055] Dynamic scenario simulation and parameter tuning are performed on the preliminary adaptive optimization control model to generate the final adaptive optimization control model. Dynamic scenario simulation needs to cover all types of working condition changes that the system may face, and comprehensively test the adaptability and stability of the preliminary model. Thirty dynamic scenarios are built using MATLAB / Simulink. In addition to a 20% increase in production load (e.g., raw material input suddenly increases from 500 tons / day to 600 tons / day), a 15% decrease in waste heat medium flow (e.g., pipeline blockage causes flow to decrease from 10 m³ / h to 8.5 m³ / h), and other scenarios such as a 5°C increase in ambient temperature (e.g., extreme high temperature in summer), equipment failure (e.g., one air compressor stops), production process switching (e.g., from mechanical processing to chemical production), and power grid voltage fluctuation (±10%), each scenario is simulated for 2 hours, and 120 time node model output data and actual optimal data are collected for comparison.

[0056] Parameter tuning is based on the results of scenario testing to improve model performance and robustness. To address the control lag and overfitting problems of the preliminary model under high load surge scenarios, the learning rate is adjusted from 0.001 to 0.0005 to slow down the parameter update speed and avoid shocks. The number of hidden layer neurons is increased from 40 / 20 to 50 / 25 to enhance the model's ability to fit complex scenarios. The L2 regularization coefficient is increased to 0.001 to suppress overfitting by penalizing large weights. The experience replay pool capacity is increased to 15,000 to include more abnormal scenario samples and improve model generalization. The target network update frequency is adjusted to every 80 steps to speed up the response to dynamic scenarios.

[0057] The final model performance needs to be verified from three dimensions: control effect, convergence speed, and stability. In the high-potential group scenario, the control effect is improved by 18% compared to the initial model. For example, during the high-temperature period in summer, the air conditioning energy consumption is reduced from 150kWh to 123kWh, and the waste heat utilization rate is increased from 65% to 76%. In the medium-potential group, the control effect is improved by 12%. In the spring and autumn normal mode, the energy consumption is reduced from 100kWh to 88kWh, and the waste heat utilization rate is increased from 75% to 84%. In the low-potential group, the control effect is improved by 8%. In the winter low-load mode, the energy consumption is reduced from 60kWh to 55.2kWh, and the waste heat utilization rate is maintained above 85%. The convergence speed is 25% faster than the initial model, and the number of training iterations is reduced from 8000 to 6000 to reach a stable state. In 20 continuous dynamic scenario tests, the model output deviation is ≤5%, which meets the requirements for stable operation of industrial systems. Finally, an adaptive optimization control model that can be directly applied to the coordinated control of industrial air conditioning and waste heat recovery is formed.

[0058] S107, based on the control strategy output of the adaptive optimization control model, combined with the dynamic correlation information of air conditioning load and waste heat supply, generates real-time control instructions and control information for energy efficiency improvement schemes.

[0059] In one implementation, control strategies for different operating modes are extracted from the final adaptive optimization control model. The strategies need to cover the target thresholds and dynamic adjustment logic of the core control factors. Examples are as follows: For the high-potential group (P1, peak summer production), the model output control strategy includes a compressor frequency target of 50Hz (±2Hz dynamically adjusted), a waste heat medium flow rate of 10m³ / h (±0.5m³ / h), heat exchangers running all day, and air conditioning supply temperature of 18℃ (±0.3℃); For the medium-potential group (P2, spring and autumn off-peak), the strategy is a compressor frequency of 40Hz (±3Hz), a waste heat medium flow rate of 8m³ / h (±0.8m³ / h), heat exchangers running intermittently according to production shifts (early shift 8:00-16:00), and supply air temperature of 19℃ (±0.5℃); For the low-potential group (P3, winter off-peak), the strategy is a compressor frequency of 30Hz (±2Hz), a waste heat medium flow rate of 6m³ / h (±0.3m³ / h), heat exchangers running only during the day shift (10:00-14:00), and supply air temperature of 20℃ (±0.5℃). Meanwhile, the strategy also includes linkage rules for auxiliary control factors, such as the condenser water temperature needing to be reduced by 0.5°C for every 1°C increase in ambient temperature, and the valve opening of the pipeline being adjusted synchronously with the fluctuation of the waste heat medium flow rate (if the flow rate decreases by 10%, the opening rate increases by 5%).

[0060] Real-time collection of air conditioning load and waste heat supply data, through correlation analysis to clear the matching relationship and dynamic change trend, to provide basis for regulation instruction correction. For example: using Internet of Things sensors to collect air conditioning cooling load every 5 minutes (such as current 190kW, 10kW higher than 10 minutes ago), waste heat generation (such as current 140kW, 8kW higher than 10 minutes ago), calculate the matching degree (140 / 190≈73.7%, lower than the model strategy set 80% target matching degree); analyze the dynamic correlation trend, find that in the past 30 minutes, the air conditioning load rises at a rate of 5kW / 10min, the waste heat supply rises at a rate of 4kW / 10min, the load grows faster than the waste heat supply, and it is predicted that the matching degree will drop to below 70% in 15 minutes; synchronous collection of environmental temperature (current 32℃, 1℃ higher than before), production load (current 85%, in the high potential group load interval), judge that the load growth is due to the increase of production process heat dissipation, and the waste heat supply growth lags behind due to the response delay of waste heat recovery equipment.

[0061] Combine dynamic correlation information with model initial strategy, dynamically correct core control factors, generate real-time control instructions that can be directly executed, and ensure the coordinated matching of air conditioning load and waste heat supply. For example: for the current high potential group scene, air conditioning load 190kW, waste heat supply 140kW (matching degree 73.7%), correct the initial strategy: compressor frequency from 50Hz to 52Hz (increase refrigerating capacity to cope with load growth), while waste heat medium flow from 10m³ / h to 11m³ / h (speed up waste heat recovery, increase supply capacity); heat exchanger remains open all day, and its heat exchange efficiency target value from 85% to 88% (achieved by increasing heat exchange area proportion); air conditioning supply air temperature remains 18℃ unchanged, but the condenser water temperature from 30℃ to 29.5℃ (offset the impact of environmental temperature rise on refrigeration efficiency); pipe valve opening degree from 80% to 85% (cooperate with waste heat medium flow to reduce transmission loss). The final real-time control instruction is output in a standardized format, including instruction identifier (such as "P1-202405201430"), control factor name, target value, execution time (such as "2024-05-2014:35 before execution"), execution mechanism (such as "compressor 1#, waste heat pump 2#, heat exchanger 1#").

[0062] Based on the expected effect of real-time regulation instructions, combined with system historical operation data and energy efficiency optimization target, an energy efficiency improvement scheme control information is formulated, including short-term adjustment, medium-term optimization and long-term improvement. For example, short-term (1-24 hours) control information: according to real-time regulation instructions, monitor air conditioning energy consumption (target from 150 kWh to below 140 kWh) and waste heat utilization rate (target from 65% to above 70%) every hour, if energy consumption is 5% higher than the target or utilization rate is 3% lower than the target, trigger secondary correction (such as further increasing waste heat medium flow by 0.5 m³ / h); Medium-term (1-7 days) control information: for high potential group scenarios, adjust production shift and time matching of waste heat recovery equipment operation (such as moving the high production peak of the middle shift from 18:00-22:00 to 1 hour earlier, synchronizing with the waste heat generation peak), the optimized target is to increase the daily waste heat utilization rate by 5% and reduce the air conditioning energy consumption by 8%; Long-term (1-3 months) control information: based on model operation data, it is suggested to maintain low-efficiency waste heat exchanger (current heat exchange efficiency is 82%, lower than the industry average of 85%), replace the aging air conditioning fan motor (current energy consumption is 10% higher than the new motor), and it is expected that the maintenance will improve the overall energy efficiency of the system by 12%, and save about 12,000 kWh of energy per year. At the same time, the scheme also includes threshold setting of energy efficiency monitoring indicators (such as triggering a warning when air conditioning energy consumption exceeds 150 kWh / hour, and triggering an alarm when waste heat utilization rate is lower than 60%) and responsible departments (such as short-term adjustment is executed by the workshop operation and maintenance group, and long-term improvement is responsible by the equipment management department).

[0063] In an embodiment, as shown in Figure 2 The present application also provides an intelligent optimization control device for industrial air conditioning and waste heat recovery cooperation, comprising:

[0064] The acquisition module 201 is used for acquiring industrial air conditioning operation parameters and waste heat recovery system data, based on multi-source data fusion technology of Internet of Things sensors and edge computing gateways, using wavelet denoising algorithm to eliminate equipment vibration interference noise, and generating standardized operation parameter sequence;

[0065] The processing module 202 is used for converting the standardized operating parameter sequence into a system energy flow network graph, and then processing by a mechanism-data hybrid modeling software, selecting a fuzzy clustering algorithm to subdivide the high energy consumption working condition, and constructing a dynamic energy efficiency coupling model; based on the dynamic energy efficiency coupling model, setting the system energy consumption as a working condition-time dependent function, using an improved particle swarm algorithm to couple a load prediction correction term to construct an optimization objective equation, extracting conversion coefficients of different waste heat qualities to construct an energy cascade utilization matrix, and performing collaborative calculation on the energy consumption distribution of the air conditioner-waste heat system to obtain key node energy loss values; combined with the key node energy loss values, extracting energy consumption peaks, waste heat utilization rates and load matching degrees according to production working condition scenes, dividing energy efficiency levels, and constructing a multi-dimensional feature matrix containing operating parameters, load characteristics and environmental condition information; based on the multi-dimensional feature matrix, comparing real-time energy efficiency with historical data to identify abnormal signals such as sudden energy consumption increase and insufficient waste heat utilization, and using the slope of the energy efficiency-load curve to generate an energy saving potential coefficient; combining the energy saving potential coefficient with the production plan to group the operation modes, using a deep reinforcement learning algorithm to screen key control factors, fusing energy efficiency parameters, equipment states and production demand information, and constructing an adaptive optimization control model; based on the output of the control strategy of the adaptive optimization control model, combining dynamic correlation information of air conditioner load and waste heat supply, generating real-time control instructions and control information of energy efficiency improvement schemes.

[0066] The computer readable storage medium provided by the above embodiments of the present application has the same beneficial effects as the method adopted, run or implemented by the application program stored therein, based on the same inventive concept as the intelligent optimization control method of the industrial air conditioner and waste heat recovery collaboration provided by the embodiments of the present application.

Claims

1. An intelligent optimization control method for industrial air conditioning and waste heat recovery cooperation, characterized in that, The method comprises the following steps: Acquire industrial air conditioner running parameters and waste heat recovery system data, based on multi-source data fusion technology of Internet of Things sensors and edge computing gateways, use wavelet denoising algorithm to eliminate equipment vibration interference noise, generate standardized running parameter sequence; After converting the standardized running parameter sequence into system energy flow network graph, processing by mechanism-data hybrid modeling software, selecting fuzzy clustering algorithm to subdivide high energy consumption conditions, build dynamic energy efficiency coupling model, including extracting and converting air conditioner energy consumption data, waste heat medium flow data, heat exchange efficiency parameters in standardized running parameter sequence, generating energy flow node association group, network transmission loss coefficient, parameter space-time matching degree; processing energy flow node association group, network transmission loss coefficient, parameter space-time matching degree, generating mechanism modeling constraint conditions, data fitting error threshold, clustering feature extraction dimension; Based on mechanism modeling constraint conditions, data fitting error threshold, clustering feature extraction dimension, classify and subdivide high energy consumption path, waste heat utilization bottleneck, condition mutation signal in system energy flow network graph, generate high energy consumption condition feature library; fuse and couple high energy consumption condition feature library, generate dynamic energy efficiency coupling model containing energy efficiency correlation equation, condition dynamic conversion rule, system coordination constraint condition; Based on dynamic energy efficiency coupling model, set system energy consumption as condition-time dependent function, use improved particle swarm algorithm to couple load prediction correction term to build optimization objective equation, extract conversion coefficient of different waste heat qualities to build energy cascade utilization matrix, perform collaborative calculation on air conditioner-waste heat system energy consumption distribution, obtain key node energy loss value; Combine key node energy loss value, extract energy consumption peak value, waste heat utilization rate, load matching degree according to production condition scene, divide energy efficiency level, build multi-dimensional feature matrix containing running parameter, load feature, environmental condition information; Based on multi-dimensional feature matrix, compare real-time energy efficiency with historical data, identify abnormal signals of energy consumption surge and insufficient waste heat utilization, use energy efficiency-load curve slope to generate energy saving potential coefficient; Combine energy saving potential coefficient with production plan, group running modes, use deep reinforcement learning algorithm to select key control factors, fuse energy efficiency parameters, equipment state and production demand information, build self-adaptive optimization control model; Output control strategy based on self-adaptive optimization control model, combine dynamic association information of air conditioner load and waste heat supply, generate real-time control instruction and energy efficiency improvement scheme control information.

2. The method of claim 1, wherein, Based on dynamic energy efficiency coupling model, set system energy consumption as condition-time dependent function, use improved particle swarm algorithm to couple load prediction correction term to build optimization objective equation, extract conversion coefficient of different waste heat qualities to build energy cascade utilization matrix, perform collaborative calculation on air conditioner-waste heat system energy consumption distribution, obtain key node energy loss value, including: Extract and adapt condition association parameters, time series energy consumption data, system coordination constraint conditions in dynamic energy efficiency coupling model, generate energy consumption function variable group, time weight coefficient, condition conversion threshold; The energy consumption function variable group, time weight coefficient and working condition conversion threshold are processed to generate an optimization target equation structure, particle swarm algorithm parameter range and load prediction correction rule; Based on the optimization target equation structure, particle swarm algorithm parameter range and load prediction correction rule, the temperature gradient, flow stability and energy conversion efficiency of different waste heat qualities are classified and quantitatively processed to generate an energy cascade utilization matrix; The energy cascade utilization matrix is subjected to collaborative calculation and iterative optimization processing to generate key node energy loss values including node instantaneous energy consumption value, cumulative loss energy and system energy efficiency proportion.

3. The method of claim 2, wherein, Combined with the key node energy loss values, the energy consumption peak value, waste heat utilization rate and load matching degree are extracted according to the production working condition scene to divide the energy efficiency level and construct a multi-dimensional feature matrix including operation parameters, load characteristics and environmental condition information, which includes: The instantaneous loss peak value, cumulative loss proportion and system energy efficiency deviation in the key node energy loss values are extracted and quantitatively processed to generate energy consumption characteristic parameters, waste heat utilization efficiency values and load matching deviation coefficients; The energy consumption characteristic parameters, waste heat utilization efficiency values and load matching deviation coefficients are processed to generate production working condition scene classification labels, energy efficiency level division thresholds and feature parameter normalization rules; Based on the production working condition scene classification labels, energy efficiency level division thresholds and feature parameter normalization rules, the operation parameter fluctuation range, load dynamic change trend and environmental condition influence weight are classified and associated to generate a multi-dimensional feature association dataset; The multi-dimensional feature association dataset is integrated and reduced to generate a multi-dimensional feature matrix including operation parameter dynamic interval, load characteristic classification vector and environmental condition influence coefficient.

4. The method of claim 1, wherein, Based on the multi-dimensional feature matrix, the real-time energy efficiency is compared with historical data to identify abnormal signals such as energy consumption surge and insufficient waste heat utilization, and the energy saving potential coefficient is generated using the energy efficiency-load curve slope, which includes: The operation parameter dynamic interval, load characteristic classification vector and environmental condition influence coefficient in the multi-dimensional feature matrix are extracted and quantitatively processed to generate real-time energy efficiency evaluation value, historical energy efficiency benchmark value and feature parameter deviation degree, wherein the historical energy efficiency benchmark value is weighted and calibrated according to the matching degree of the same period production working condition and environmental condition; The real-time energy efficiency evaluation value, historical energy efficiency benchmark value and feature parameter deviation degree are processed to generate energy efficiency comparison difference value, abnormal signal identification threshold and fluctuation trend judgment rule, and the abnormal signal identification threshold is set to a dynamic threshold interval according to the production load intensity classification; Based on the energy efficiency comparison difference value, abnormal signal identification threshold and fluctuation trend judgment rule, the energy consumption surge amplitude, waste heat utilization gap and energy efficiency-load curve form are classified and slope calculation processed to generate an energy saving potential evaluation dataset, and the slope calculation is corrected by introducing the attenuation coefficient of equipment operation life; The energy saving potential evaluation dataset is integrated and quantitatively converted to generate an energy saving potential coefficient including instantaneous energy saving space, cumulative energy saving potential and dynamic adjustment coefficient, and the dynamic adjustment coefficient is dynamically corrected in association with the real-time load rate of the waste heat recovery equipment.

5. The method of claim 1, wherein, The energy-saving potential coefficient is combined with the production plan, the operation modes are grouped, the deep reinforcement learning algorithm is used to screen the key regulation factors, the energy efficiency parameters, the equipment state and the production demand information are fused, and a self-adaptive optimization control model is constructed, including: In combination with the production plan, key production parameters affecting the system operation mode are extracted, and the operation mode is divided into a high potential group, a medium potential group and a low potential group according to the energy efficiency optimization level, wherein each group of modes corresponds to a unique optimization priority label; The energy-saving potential coefficients, the energy efficiency parameters and the production demand information of different potential groups are compared and analyzed to generate characteristic difference data between groups; The characteristic difference data, the energy-saving potential coefficient, the energy efficiency parameter, the equipment state and the production demand information are integrated and processed to generate a multi-dimensional model training data set, wherein the energy efficiency parameter includes a key node energy loss value and a waste heat utilization rate; Based on the multi-dimensional model training data set, the deep reinforcement learning algorithm is used to screen out key regulation factors that have a significant impact on the control effect, and a preliminary self-adaptive optimization control model including a strategy network and a value network is constructed; The preliminary self-adaptive optimization control model is simulated and parameterized in a dynamic scene to generate a final self-adaptive optimization control model.

6. An intelligent optimization control device for industrial air conditioning and waste heat recovery, characterized in that, The device for implementing the method of claim 1 comprises: A collection module is used to acquire industrial air conditioner operation parameters and waste heat recovery system data, and based on the multi-source data fusion technology of Internet of Things sensors and edge computing gateways, a wavelet denoising algorithm is used to eliminate device vibration interference noise to generate a standardized operation parameter sequence; A processing module is used to convert the standardized operation parameter sequence into a system energy flow network graph, and then process it through a mechanism-data hybrid modeling software, select a fuzzy clustering algorithm to subdivide high energy consumption conditions, and construct a dynamic energy efficiency coupling model; based on the dynamic energy efficiency coupling model, set the system energy consumption as a condition-time dependent function, use an improved particle swarm algorithm to couple a load prediction correction term to construct an optimization objective equation, extract conversion coefficients of different waste heat qualities to construct an energy cascade utilization matrix, and perform collaborative calculation on the air conditioner-waste heat system energy consumption distribution to obtain a key node energy loss value; in combination with the key node energy loss value, extract the energy consumption peak value, the waste heat utilization rate and the load matching degree according to the production condition scene, divide the energy efficiency level, and construct a multi-dimensional feature matrix containing operation parameter, load feature and environmental condition information; based on the multi-dimensional feature matrix, compare the real-time energy efficiency with the historical data to identify abnormal signals such as energy consumption surge and insufficient waste heat utilization, and use the energy efficiency-load curve slope to generate an energy-saving potential coefficient; the energy-saving potential coefficient is combined with the production plan to group the operation modes, the deep reinforcement learning algorithm is used to screen the key regulation factors, the energy efficiency parameters, the equipment state and the production demand information are fused, and a self-adaptive optimization control model is constructed; based on the regulation strategy output of the self-adaptive optimization control model, in combination with the dynamic correlation information of the air conditioner load and the waste heat supply, control information of real-time regulation instructions and energy efficiency improvement schemes is generated.

7. An electronic device, comprising: It comprises: A first processor; And a memory for storing executable instructions of the first processor; The first processor is configured to execute the industrial air conditioner and waste heat recovery collaborative intelligent optimization control method according to any one of claims 1-5 by executing the executable instructions.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the second processor to implement the industrial air conditioner and waste heat recovery collaborative intelligent optimization control method according to any one of claims 1-5.

Images