Central air conditioner cooling tower control method based on side cloud cooperative computing

Abstract

The invention discloses a central air conditioner cooling tower control method based on edge cloud cooperative computing, and belongs to the technical field of cooling towers. Data from a sensor is collected and transmitted to an edge node for primary processing and fusion; performing optimization processing on the acquired data by using an improved particle swarm algorithm; then optimizing the water distribution flow by adopting a beetle antennae-particle swarm fusion optimization algorithm; and finally realizing intelligent control of the cooling tower cluster through a method of combining deep reinforcement learning and model prediction control. By the adoption of the central air conditioner cooling tower control method based on side cloud cooperative computing, the intelligent degree and energy efficiency of the cooling tower can be improved, energy consumption is reduced, and the cooling effect is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of cooling towers, and specifically relates to a central air-conditioning cooling tower control method based on edge-cloud collaborative computing. Background Art

[0002] Cooling towers are one of the most important equipment in refrigeration systems and are widely used in data centers, thermal power plants, buildings and other fields. At present, the challenge of high energy consumption of industrial cooling towers at home and abroad has not been systematically studied and effectively solved, and the current situation of low automation and high economic consumption also needs to be improved. Therefore, with the market development of data centers and the energy industry, the demand for cooling towers and their cooling systems has gradually increased, and China urgently needs to develop a new type of intelligent, efficient and energy-saving cooling tower. At the same time, with the development of cutting-edge technologies such as artificial intelligence Internet of Things, cloud computing and artificial intelligence in China, the adjustment of the low degree of intelligence of cooling towers has become increasingly prominent.

[0003] In terms of edge-cloud collaboration, cooling tower control systems face many challenges: low performance and single functions of edge node devices make it difficult to process large amounts of time series data; single systems lack high-availability recovery mechanisms and are susceptible to node failures; the large amount of data generated by multiple sensors exceeds the communication load capacity of a single cloud architecture, leading to network pressure and latency issues; and the real-time nature of cloud computing makes it difficult to meet the high demands of cooling tower cluster control. At the same time, traditional cooling tower clusters have poor results in data fusion and transmission, and particle swarm algorithms have deficiencies in practical applications. Excessive energy consumption and uneven water distribution during operation lead to poor cooling effects. Therefore, a new technical solution is urgently needed. Summary of the invention

[0004] The purpose of the present invention is to provide a central air-conditioning cooling tower control method based on edge-cloud collaborative computing, which can improve the intelligence and energy efficiency of the cooling tower, reduce energy consumption, and improve the cooling effect.

[0005] To achieve the above object, the present invention provides a central air-conditioning cooling tower control method based on edge-cloud collaborative computing, comprising the following steps:

[0006] S1. Build a cloud-edge-end system. The cloud-edge-end system includes the device layer, edge computing layer, and cloud layer. The device layer includes the intelligent sensor module.

[0007] S2, the data collection layer uses the intelligent sensor module in S1 to collect real-time data. After the edge computing nodes in the edge computing layer receive the real-time data, they use the improved particle swarm algorithm to process the real-time data and save the processed data in the edge computing data pool;

[0008] S3, using a method combining deep reinforcement learning and model predictive control to achieve intelligent control of cooling tower clusters;

[0009] S4, using beetle whisker-particle swarm fusion optimization algorithm to control water distribution flow;

[0010] S5. Execute S2 to S4 in a loop, and continuously adjust the operation mode of the cooling tower according to real-time data and historical records.

[0011] Preferably, the improved particle swarm algorithm in S2 is calculated as follows:

[0012]

[0013] Among them, ω represents the inertia weight coefficient, ω(j) represents the inertia weight coefficient at the current iteration step j, and ω max represents the maximum inertia weight coefficient, ω min represents the minimum inertia weight coefficient, T max represents the maximum number of iterations, j represents the current iteration step;

[0014] c1(j)=c 11 -(c 11 -c 12 )×(j / T max );

[0015] c2(j)=c 21 -(c 22 -c 21 )×(j / T max );

[0016] Where c1(j) represents the acceleration factor c1 at the current iteration step j, c2(j) represents the acceleration factor c2 at the current iteration step j, and c 11 and c 21 represents the initial value, c 12 and c 22 Indicates the termination value;

[0017] The inertia weight coefficient and acceleration factor are integrated to update the particle's velocity and position. The particle velocity and position are updated as follows:

[0018] v id =ω(j)v id +c1(j)r1(P id -X id )+c2(j)r2(C id -X id );

[0019] Among them, v id represents the velocity of particle i in dimension d, P id represents the historical optimal position of particle i in dimension d, X id represents the current position of particle i in dimension d, Cid represents the value of the global optimal position in dimension d, r1 and r2 represent two independent and identically distributed random numbers in [0,1], c1 and c2 represent acceleration factors, which are usually set to 2;

[0020] X i (d+1)=X id +V id ;

[0021] Among them, X i (d+1) represents the new position of particle i in dimension d+1, V id represents the new velocity of particle i in dimension d.

[0022] Preferably, in S3, the method based on combining deep reinforcement learning with model predictive control includes the following steps:

[0023] S301, set initial parameters, define reward function, prepare training environment, initialize SAC algorithm agent and MPC controller;

[0024] S302, the SAC algorithm agent generates an optimal control parameter set Δu according to the current state of the cooling tower cluster DRL ; The MPC controller receives the optimal control parameter set Δu output by the SAC algorithm agent DRL ; The MPC controller is based on Δu DRL Calculate the specific control signal Δu MPC ; The MPC controller is based on Δu MPC , sends an actual load frequency control signal to the frequency conversion unit in the cooling tower, and the load frequency control signal is used to adjust the operating parameters of the frequency conversion unit;

[0025] S303, the SAC algorithm agent gradually improves its decision-making process, making the system tend to a more energy-saving and efficient operation mode.

[0026] Preferably, in S4, the beetle whisker-particle swarm algorithm comprises the following steps:

[0027] S401, setting parameters and initializing particles, where the particles represent the water distribution flow configuration scheme;

[0028] S402: Calculate the fitness function value f(x) according to the water distribution flow configuration represented by the particle, and determine the individual position P corresponding to the best fitness in the particle group t and the global optimal position P k ;

[0029] S403, according to P t and P k The information is used to adjust the particle's speed and position. The calculation method is as follows:

[0030]

[0031] in, represents the position of particle i in dimension d at the kth iteration, represents the velocity of particle i in dimension d at the kth iteration, represents the historical optimal solution of particle i in dimension d, Represents the value of the global optimal position in dimension d;

[0032] S404, resetting parameters using the updated particle positions, calculating the total energy consumption under the new water distribution flow configuration, and evaluating whether the data solution for optimal energy efficiency is met, and using the following formula to update the parameters:

[0033]

[0034] in, Indicates the water distribution flow or water outflow on the left side. Indicates the water distribution flow or water outflow on the right side, x k represents the current position or flow rate, d0 represents the step size, and dir represents a binary variable, which usually takes a value of +1 or -1;

[0035] Calculate the direction vector sign to determine the direction of movement:

[0036]

[0037] Update the example position according to the direction vector sign and step size step:

[0038] x k+1 =x k -dir·step·sign;

[0039] S405. If the new configuration reaches the preset optimization standard, the optimized data is saved; otherwise, the optimization process is continued by returning to S402 until the convergence condition is reached or the predetermined number of iterations is completed;

[0040] S406. When the optimization process is finished, the water distribution flow parameters are output.

[0041] Preferably, the real-time data includes water inlet water temperature information, water outlet water temperature information, turbine water inlet flow rate information, fan speed information, water distribution flow information, cooling tower heat transfer information and water pump water pressure information.

[0042] Preferably, the cloud layer is used to provide computing and storage services; the edge computing layer is used to manage the cooling tower cluster and optimize local data processing and data transmission in the edge network.

[0043] Preferably, the cloud layer module includes a cloud computing module, a cloud storage module and a cloud platform module. The cloud computing module provides computing services, the cloud storage module is used to store data uploaded from edge nodes, and the cloud platform module is used to manage user access rights, task scheduling and service coordination, and provides a standardized API interface.

[0044] Preferably, the edge computing layer includes an execution control module, a data perception module, a modeling and analysis module, and an intelligent decision-making module.

[0045] Preferably, the equipment layer also includes a PLC equipment module and a cooling tower body module.

[0046] Therefore, the present invention adopts the above-mentioned central air-conditioning cooling tower control method based on edge-cloud collaborative computing. Compared with the prior art, the present invention has the following significant beneficial effects:

[0047] (1) The present invention realizes the control of water distribution flow, avoids the blockage problem, improves the efficiency of water resource utilization, and increases the heat transfer efficiency of the cooling tower;

[0048] (2) The present invention constructs an improved particle swarm algorithm to perform data processing at the edge node and optimize the cooling tower operating parameters, thereby effectively reducing the total power consumption of the entire cooling tower system and achieving optimal energy efficiency;

[0049] (3) The present invention reduces the reliance on manual intervention, reduces manual operation and maintenance costs, and improves operation and maintenance efficiency;

[0050] (4) The present invention proposes a four-layer structure of structural equipment layer, data collection layer, edge computing layer and cloud layer, and uses the powerful computing power of edge computing close to the equipment structure for real-time data fusion of cooling tower clusters;

[0051] (5) The architecture proposed in the present invention has the advantages of fast information transmission and strong compatibility. It supports multiple cooling towers to share cloud service resources and adapt to different business needs. It can also expand and control more cooling towers by adding PCI.

[0052] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a technical architecture diagram of a central air-conditioning cooling tower control method based on edge-cloud collaborative computing of the present invention;

[0054] Figure 2 This is an edge node-cloud system framework design diagram of a central air-conditioning cooling tower control method based on edge-cloud collaborative computing of the present invention;

[0055] Figure 3This is a cloud-edge-end system framework diagram of a central air-conditioning cooling tower control method based on edge-cloud collaborative computing of the present invention;

[0056] Figure 4 A real-time data processing flow chart of a central air-conditioning cooling tower control method based on edge-cloud collaborative computing of the present invention;

[0057] Figure 5 This is a flow chart of an improved particle swarm algorithm for a central air-conditioning cooling tower control method based on edge-cloud collaborative computing of the present invention;

[0058] Figure 6 A schematic diagram of cooling tower cluster control of a central air-conditioning cooling tower control method based on edge-cloud collaborative computing of the present invention;

[0059] Figure 7 This is a flow chart of a water distribution flow optimization algorithm for a central air-conditioning cooling tower control method based on edge-cloud collaborative computing of the present invention;

[0060] Figure 8 This is a flowchart of the actual operation of a water distribution flow optimization algorithm of a central air-conditioning cooling tower control method based on edge-cloud collaborative computing in the present invention;

[0061] Fig. 9 This is a cooling tower improved particle swarm algorithm data diagram of a central air-conditioning cooling tower control method based on edge-cloud collaborative computing of the present invention;

[0062] Fig.10 This is a data graph of the energy saving rate of the cooling tower after optimization of a central air-conditioning cooling tower control method based on edge-cloud collaborative computing in the present invention. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Unless otherwise defined, the technical terms or scientific terms used in the present invention should be the common meaning understood by people with general skills in the field to which the present invention belongs.

[0064] Embodiment 1

[0065] like Figure 1-Figure 2 As shown, a central air-conditioning cooling tower control method based on edge-cloud collaborative computing of the present invention comprises the following steps:

[0066] S1, such as Figure 3As shown in the figure, a cloud-edge-end system is constructed, which includes the device layer, edge computing layer and cloud layer. The system is divided into reasonable modules to meet the problem handling and function realization encountered in the actual operation and maintenance of cooling towers.

[0067] The equipment layer includes PLC equipment module, cooling tower body module and intelligent sensor module. The PLC equipment module includes communication cards responsible for motion control, I / O communication cards for auxiliary system functions, data collection module cards and other PCI function cards, which are transferred to the execution end of the cooling tower through the terminal block, and support the expansion control of multiple cooling towers by adding PCI. The cooling tower body module contains a water pump motor and corresponding driver for water delivery, as well as auxiliary equipment such as water replenishment devices and drainage devices that perform auxiliary functions. This module can be selectively linked to the industrial computer according to the required functions to complete the required cooling tower command control and intelligent function execution. The intelligent sensor module includes temperature sensors, pressure sensors, flow sensors, etc., which are used to obtain various parameters in the operation of the cooling tower.

[0068] The structural equipment layer is composed of the cooling tower body module, which is used to receive the control instructions from the edge layer. It adjusts the operating parameters of the internal equipment of the cooling tower (such as water pump speed, fan speed, etc.) according to the control instructions to perform specific cooling tasks.

[0069] The data acquisition layer uses intelligent sensor modules to collect inlet water temperature information, outlet water temperature information, turbine inlet flow rate information, fan speed information, water distribution flow information, cooling tower heat transfer information, and water pump water pressure information, and sends the above real-time data to the edge computing layer through the transmission layer.

[0070] The transport layer uses the MQTT protocol for data push based on the application layer transmission of TCP connection, and 5G is used for real-time data transmission at the physical layer / data link layer.

[0071] The edge computing layer is composed of multiple edge computing nodes, which are distributed near each cooling tower to form a distributed computing network. The edge computing layer includes data perception module, modeling and analysis module, intelligent decision-making module and control execution module. The data perception module is responsible for information intelligent scheduling and data preprocessing, including data cleaning, time alignment, feature extraction, association fusion, intelligent labeling and other processes, and saves the processed data in the edge computing data pool to provide guarantee for data extraction and analysis calls of other modules. The modeling and analysis module extracts the perceived data information from the edge computing data pool, and combines the mathematical model of theoretical research to analyze and construct the cooling tower status (such as temperature status, pressure status, flow status, etc.) under different operating scenarios. The intelligent decision-making module forms information and instructions to support decision-making, and realizes real-time monitoring, data analysis, optimization control and intelligent decision-making of the operation process of the cooling tower cluster. The control execution module outputs control instructions to the underlying equipment and receives data feedback from the underlying equipment to form a closed-loop control.

[0072] The application layer is used to establish the cloud layer to analyze and apply the data uploaded by the edge nodes. The cloud layer includes cloud computing module, cloud storage module and cloud platform module. The cloud computing module provides powerful computing services. When faced with large amounts of data computing or simulation analysis needs, cloud computing resources are remotely called to complete the analysis or decision-making tasks of the cooling tower system. The cloud storage module is used to store various types of data uploaded from edge nodes, such as historical data collected by sensors, optimized parameter configurations, etc. This module supports functions such as long-term trend analysis, historical data review and fault diagnosis, providing a basis for subsequent data mining and deep learning. The cloud platform module manages user access rights, task scheduling and service coordination, provides standardized API interfaces, supports multiple tenants to use the same set of cloud platform services, and ensures the stable operation of services.

[0073] The above-mentioned structural equipment layer, data collection layer, edge computing layer and cloud layer constitute the edge node-cloud system framework, which uses the powerful computing power of edge computing close to the equipment structure for real-time data fusion of cooling tower clusters.

[0074] S2, such as Figure 4-Figure 5As shown, the data acquisition layer uses the intelligent sensor module in S1 to collect real-time data. After the edge computing nodes in the edge computing layer receive the real-time data, the improved particle swarm algorithm is used to optimize the data at the edge nodes. First, the speed and position of the particles are initialized, and the real-time data is input, including the water temperature information of the water inlet, the water temperature information of the water outlet, the water inlet flow rate information of the turbine, the fan speed information, the water distribution flow information, the cooling tower heat transfer information and the water pump water pressure information. After preprocessing the input real-time data, the total power consumption of the cooling tower system in the current state is calculated based on the preprocessed data, and the position of each particle corresponds to the performance index, inertia weight coefficient and acceleration factor of the system. The calculation method is as follows:

[0075]

[0076] Among them, ω represents the inertia weight coefficient, ω(j) represents the inertia weight coefficient at the current iteration step j, and ω max represents the maximum inertia weight coefficient, ω min represents the minimum inertia weight coefficient, T max represents the maximum number of iterations, j represents the current iteration step;

[0077] c1(j)=c 11 -(c 11 -c 12 )×(j / T max );

[0078] c2(j)=c 21 -(c 22 -c 21 )×(j / T max );

[0079] Where c1(j) represents the acceleration factor c1 at the current iteration step j, c2(j) represents the acceleration factor c2 at the current iteration step j, and c 11 and c 21 represents the initial value, c 12 and c 22 Indicates the end value.

[0080] The inertia weight coefficient and acceleration factor are integrated to update the particle's velocity and position. The particle velocity and position are updated as follows:

[0081] v id =ω(j)v id +c1(j)r1(P id -X id )+c2(j)r2(C id -X id );

[0082] Among them, v idrepresents the velocity of particle i in dimension d, P id represents the historical optimal position of particle i in dimension d, X id represents the current position of particle i in dimension d, C id represents the value of the global optimal position in dimension d, r1 and r2 represent two independent and identically distributed random numbers in [0,1], c1 and c2 represent acceleration constants, which are usually set to 2;

[0083] X i (d+1)=X id +V id ;

[0084] Among them, X i (d+1) represents the new position of particle i in dimension d+1, V id represents the new velocity of particle i in dimension d.

[0085] By continuously adjusting the speed and position of particles, the improved particle swarm algorithm can effectively explore the solution space and find the optimal solution. Then, the total energy consumption of the cooling tower is calculated again using the updated particle position to find new local optimal values ​​and global optimal values ​​to determine whether a better solution has been achieved. If the newly calculated total energy consumption meets the preset optimization criteria, the optimized data is uploaded to the cloud and saved in the edge computing data pool. If the conditions are not met, the optimization process continues until a satisfactory optimization result is achieved or the predetermined number of iterations is completed.

[0086] S3, such as Figure 6 As shown, in order to realize the intelligent control of cooling tower cluster, the present invention adopts a method combining deep reinforcement learning (DRL) and model predictive control (MPC), in which deep reinforcement learning adopts SAC (Soft Actor-Critic) algorithm. Through this combination, the parameters of the model predictive controller established in the cooling tower can be adaptively adjusted to improve energy utilization efficiency and reduce energy consumption. The following is the specific workflow of the method:

[0087] S301, set initial parameters, define reward function, prepare training environment, initialize SAC algorithm agent and MPC controller;

[0088] S302, the SAC algorithm agent generates an optimal control parameter set Δu according to the current state of the cooling tower cluster (including but not limited to the variable flow frequency fluctuation set and the control signal information of other cooling tower controllers) DRL, these parameters are based on the strategy obtained by the SAC algorithm agent through continuous learning and optimization; the MPC controller receives the optimal control parameter set Δu output by the SAC algorithm agent DRL Based on these parameters, the MPC controller calculates the specific control signal Δu MPC ; The MPC controller is based on Δu MPC , sends actual load frequency control (LFC) signals to the variable frequency units in each cooling tower. These LFC signals are used to adjust the actual operating parameters of the variable frequency units and generate changes, so as to achieve collaborative control of the cooling tower cluster. After each cooling tower executes the control instruction, it obtains a new state observation and immediate reward (or penalty), which reflects the effect of this control. The SAC algorithm agent updates its internal strategy based on the new state and reward so as to make better choices in the future;

[0089] S303. With the accumulation of time and experience, the SAC algorithm agent gradually improves its decision-making process, making the system tend to a more energy-saving and efficient operation mode.

[0090] The present invention combines deep reinforcement learning with model predictive control. The SAC algorithm agent generates optimal control parameters based on real-time environmental information, and the MPC controller receives these parameters and generates specific LFC signals, thereby achieving coordinated and stable operation of the cooling tower cluster. This method is suitable for complex industrial environments that require high precision and dynamic adjustment, such as central air conditioning cooling systems in data centers, large building complexes, and other places.

[0091] S4. In order to achieve water distribution flow control, the beetle whisker-particle swarm algorithm is used to optimize the water distribution flow. The beetle whisker-particle swarm algorithm is used to set system parameters, calculate fitness and update the P corresponding to the best fitness of the parameters. t and P k value, and further adjust the water distribution flow and other related parameters.

[0092] like Figure 7-Figure 8 As shown in the figure, Beetle Antennae Search-Particle Swarm Optimization (BAS-PSO) is a hybrid algorithm that integrates biological heuristic optimization and is used to solve the water distribution flow optimization problem in central air-conditioning cooling towers to achieve uniform water distribution. The following are the specific application steps of the algorithm in setting system parameters, calculating fitness and updating the position corresponding to the optimal fitness, and adjusting the water distribution flow and other related parameters:

[0093] S401, setting system parameters such as inlet and outlet water flow, inlet water flow, and initializing a group of particles, each particle representing a possible water distribution flow configuration scheme;

[0094] S402. For each particle, calculate the fitness function value f(x) according to the water distribution flow configuration it represents. The fitness function is designed to be a comprehensive indicator reflecting goals such as water distribution uniformity and energy saving effect. Determine the individual position P corresponding to the best fitness in the current particle group t (i.e. the optimal position of the current particle in the previous n iterations) and the global optimal position P k (i.e. the optimal position of all particles in the first n iterations).

[0095] S403, according to P t and P k The information is used to adjust the speed and position of the particles using the beetle whisker search mechanism to further optimize the water distribution flow configuration. The calculation method is as follows:

[0096]

[0097] in, represents the position of particle i in dimension d at the kth iteration, represents the velocity of particle i in dimension d at the kth iteration, represents the historical optimal solution of particle i in dimension d, Represents the value of the global optimal position in dimension d;

[0098] S404. Use the updated particle position to reset the water distribution flow and other related parameters (such as the water outlet flow, etc.), calculate the total energy consumption under the new water distribution flow configuration, and evaluate whether the data solution for optimal energy efficiency is met. Use the following formula to update the water distribution flow, water outlet flow and other parameters:

[0099]

[0100] in, Indicates the water distribution flow or water outflow on the left side. Indicates the water distribution flow or water outflow on the right side, x k represents the current position or flow rate, d0 represents the step size, and dir represents a binary variable, which usually takes a value of +1 or -1;

[0101] Calculate the direction vector sign to determine the direction of movement:

[0102]

[0103] Update the example position according to the direction vector sign and step size step:

[0104] x k+1 =x k -dir·step·sign;

[0105] S405. If the new configuration reaches the preset optimization standard, the optimized data is saved; otherwise, the optimization process is continued by returning to S402 until the convergence condition is reached or the predetermined number of iterations is completed;

[0106] S406. When the optimization process is finished, the optimal water distribution flow parameters finally determined are output to ensure that the water distribution during the operation of the cooling tower is more uniform and reasonable, reduce the drift rate, and improve the water resource utilization efficiency and heat transfer efficiency.

[0107] S5. Execute S2 to S4 in a loop, and continuously adjust the operation mode of the cooling tower according to real-time data and historical records to achieve the best energy efficiency ratio.

[0108] like Figure 9-10 As shown, through the optimization measures, the cooling tower shows a high energy saving rate during most of the day, which proves the effectiveness of the present invention.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A central air conditioning cooling tower control method based on edge-cloud collaborative computing, characterized in that: The following steps are involved: S1. Build a cloud-edge-end system. The cloud-edge-end system includes the device layer, edge computing layer, and cloud layer. The device layer includes the intelligent sensor module. S2, the data collection layer uses the intelligent sensor module in S1 to collect real-time data. After the edge computing nodes in the edge computing layer receive the real-time data, they use the improved particle swarm algorithm to process the real-time data and save the processed data in the edge computing data pool; S3, using a method combining deep reinforcement learning and model predictive control to achieve intelligent control of cooling tower clusters; S4, using beetle whisker-particle swarm fusion optimization algorithm to control water distribution flow; S5. Execute S2 to S4 in a loop, and continuously adjust the operation mode of the cooling tower according to real-time data and historical records.

2. A central air conditioning cooling tower control method based on edge-cloud collaborative computing according to claim 1, characterized in that: The improved particle swarm algorithm in S2 is calculated as follows: Among them, ω represents the inertia weight coefficient, ω(j) represents the inertia weight coefficient at the current iteration step j, and ω max represents the maximum inertia weight coefficient, ω min represents the minimum inertia weight coefficient, T max represents the maximum number of iterations, j represents the current iteration step; c1(j)=c 11 -(c 11 -c 12 )×(j / T max ); c2(j)=c 21 -(c 22 -c 21 )×(j / T max ); Where c1(j) represents the acceleration factor c1 at the current iteration step j, c2(j) represents the acceleration factor c2 at the current iteration step j, and c 11 and c 21 represents the initial value, c 12 and c 22 Indicates the termination value; The inertia weight coefficient and acceleration factor are integrated to update the particle's velocity and position. The particle velocity and position are updated as follows: v id =ω(j)v id +c1(j)r1(P id -X id )+c2(j)r2(C id -X id ); Among them, v id represents the velocity of particle i in dimension d, P id represents the historical optimal position of particle i in dimension d, X id represents the current position of particle i in dimension d, C id represents the value of the global optimal position in dimension d, r1 and r2 represent two independent and identically distributed random numbers in [0,1], c1 and c2 represent acceleration factors, which are usually set to 2; X i (d+1)=X id +V id ; Among them, X i (d+1) represents the new position of particle i in dimension d+1, V id represents the new velocity of particle i in dimension d.

3. A central air conditioning cooling tower control method based on edge-cloud collaborative computing according to claim 1, characterized in that: In S3, the method based on combining deep reinforcement learning with model predictive control includes the following steps: S301, set initial parameters, define reward function, prepare training environment, initialize SAC algorithm agent and MPC controller; S302, the SAC algorithm agent generates an optimal control parameter set Δu according to the current state of the cooling tower cluster DRL ; The MPC controller receives the optimal control parameter set Δu output by the SAC algorithm agent DRL ; The MPC controller is based on Δu DRL Calculate the specific control signal Δu MPC ; The MPC controller is based on Δu MPC , sends an actual load frequency control signal to the frequency conversion unit in the cooling tower, and the load frequency control signal is used to adjust the operating parameters of the frequency conversion unit; S303, the SAC algorithm agent gradually improves its decision-making process, making the system tend to a more energy-saving and efficient operation mode.

4. A central air conditioning cooling tower control method based on edge-cloud collaborative computing according to claim 1, characterized in that: In S4, the beetle whisker-particle swarm algorithm includes the following steps: S401, setting parameters and initializing particles, where the particles represent the water distribution flow configuration scheme; S402: Calculate the fitness function value f(x) according to the water distribution flow configuration represented by the particle, and determine the individual position P corresponding to the best fitness in the particle group t and the global optimal position P k ; S403, according to P t and P k The information is used to adjust the particle's speed and position. The calculation method is as follows: in, represents the position of particle i in dimension d at the kth iteration, represents the velocity of particle i in dimension d at the kth iteration, represents the historical optimal solution of particle i in dimension d, Represents the value of the global optimal position in dimension d; S404, resetting parameters using the updated particle positions, calculating the total energy consumption under the new water distribution flow configuration, and evaluating whether the data solution for optimal energy efficiency is met, and using the following formula to update the parameters: in, Indicates the water distribution flow or water outflow on the left side. Indicates the water distribution flow or water outflow on the right side, x k represents the current position or flow rate, d0 represents the step size, and dir represents a binary variable, which usually takes a value of +1 or -1; Calculate the direction vector sign to determine the direction of movement: Update the example position according to the direction vector sign and step size step: x k+1 =x k -dir·step·sign; S405. If the new configuration reaches the preset optimization standard, the optimized data is saved; otherwise, the optimization process is continued by returning to S402 until the convergence condition is reached or the predetermined number of iterations is completed; S406. When the optimization process is finished, the water distribution flow parameters are output.

5. A central air conditioning cooling tower control method based on edge-cloud collaborative computing according to claim 1, characterized in that: Real-time data includes water inlet temperature information, water outlet temperature information, turbine inlet flow rate information, fan speed information, water distribution flow information, cooling tower heat transfer information and water pump water pressure information.

6. A central air conditioning cooling tower control method based on edge-cloud collaborative computing according to claim 1, characterized in that: The cloud layer is used to provide computing and storage services; the edge computing layer is used to manage cooling tower clusters and optimize local data processing and data transmission on edge networks.

7. A central air conditioning cooling tower control method based on edge-cloud collaborative computing according to claim 1, characterized in that: The cloud layer module includes cloud computing module, cloud storage module and cloud platform module. The cloud computing module provides computing services, the cloud storage module is used to store data uploaded from edge nodes, and the cloud platform module is used to manage user access rights, task scheduling and service coordination, and provides a standardized API interface.

8. A central air conditioning cooling tower control method based on edge-cloud collaborative computing according to claim 1, characterized in that: The edge computing layer includes an execution control module, a data perception module, a modeling and analysis module, and an intelligent decision-making module.

9. A central air conditioning cooling tower control method based on edge-cloud collaborative computing according to claim 1, characterized in that: The equipment layer also includes PLC equipment module and cooling tower body module.

Images