Air cooling island anti-freezing control method and system based on AI prediction and two-stage optimization

Abstract

The embodiment of the invention provides an air cooling island anti-freezing control method and system based on AI prediction and dual-stage optimization, and the method is characterized in that the method comprises the steps: carrying out the analysis of collected air cooling island data through an AE model, evaluating the freezing probability of each fan region in a given time window, and calculating the freezing probability of each fan region; according to the freezing probability, the freezing risk of each fan area is divided into three levels of low risk, medium risk and high risk; when the freezing risk is identified as a medium risk or a high risk, establishing a first optimization model of fan speed change and temperature change of a corresponding temperature measurement zone, and performing first-stage optimization by taking air cooling island freezing risk minimization as a target; when the first optimization model cannot meet the anti-freezing requirement, establishing a second optimization model of the rotating speed of the reverse fan, the start and stop of the vacuum pump and the temperature of the temperature measuring zone, and performing second-stage optimization; the optimization result is converted into a control instruction capable of being issued, and feedback signals are recycled to correct the automatic encoder model in a closed loop mode.

Description

Technical Field

[0001] This document relates to the field of energy and power technology, and in particular to a method and system for antifreeze control of air-cooled islands based on AI prediction and two-stage optimization. Background Technology

[0002] An air-cooled condenser (ACC) is a cooling system that uses air as the cooling medium to condense steam discharged from a steam turbine. It is widely used in energy systems such as thermal power plants and combined heat and power (CHP) systems. Compared to traditional water-cooled systems, air-cooled condensers have advantages such as lower water consumption and suitability for arid or cold regions.

[0003] Air-cooled islands typically consist of multiple fan units, each containing heat exchange tube bundles, fans, a transmission system, and control components. Ambient air is drawn in by the fans, exchanging heat with steam to form condensate, which is then recycled in a closed loop. As the terminal unit of the entire condensation system, the air-cooled island's operating status significantly impacts the unit's safety, thermal efficiency, and load regulation capabilities. In low-temperature winter environments, especially in northern my country and high-altitude areas, air-cooled islands face a severe risk of freezing. The heat exchanger surfaces are prone to icing under conditions of low temperature, strong winds, and humidity, potentially leading to severe freezing blockage, affecting heat exchange efficiency, and in extreme cases, causing equipment damage or unplanned shutdowns of the entire unit, resulting in serious safety hazards and economic losses.

[0004] To address icing issues in cold weather, current mainstream anti-freezing measures include: shutting down some fans to reduce cold air inflow; increasing exhaust steam pressure or raising steam-side temperature; using bypass steam for regional heating; installing anti-freezing electric heating devices; and manual inspection and adjustment of fan status. Most of these methods rely on static, experience-based judgment or rule-based control, lacking real-time identification and response to actual freezing risks. Especially under extreme weather conditions, these strategies often fail to accurately and effectively prevent localized freezing, leading to frequent start-ups and shutdowns, increased energy consumption, and even equipment failure. Existing control systems prioritize "freezing prevention" without considering energy consumption indicators, often resulting in large-scale fan shutdowns or bypass steam causing significant energy losses and low operating efficiency. Furthermore, due to a lack of comprehensive judgment on future weather trends, and the fact that most current systems can only react after freezing has occurred, the optimal window for early prevention is missed. Summary of the Invention

[0005] This specification provides one or more embodiments of an air-cooled island antifreeze control method based on AI prediction and two-stage optimization, including:

[0006] The collected air-cooled island data is analyzed using the automatic encoder model AE to assess the freezing probability of each wind turbine area within a given time window, and the freezing risk of each wind turbine area is divided into three levels: low risk, medium risk, and high risk based on the freezing probability.

[0007] When the freezing risk is identified as medium or high, a first optimization model is established to correlate the fan speed change with the corresponding temperature change of the temperature measurement zone. The first stage of optimization is carried out with the goal of minimizing the freezing risk of the air-cooled island.

[0008] When the first optimization model cannot meet the antifreeze requirements, a second optimization model is established, which includes the reverse fan speed, vacuum pump start / stop and temperature measurement zone temperature. The second stage of optimization is carried out with the goal of minimizing the risk of freezing in the air-cooled island.

[0009] The optimization results are converted into control commands that can be issued, and feedback signals are collected to perform closed-loop correction of the automatic encoder model.

[0010] Furthermore, data on the air-cooled island is acquired from the DCS system and edge sensors. This data includes: condensate temperature, temperature of the air-cooled island surface temperature measuring zone, ambient temperature, ambient humidity, fan speed, wind speed, wind direction, main steam flow rate, total coal consumption, total air volume, low-pressure cylinder inlet steam pressure, low-pressure cylinder inlet steam temperature, condensate pump outlet header temperature, temperature and pressure of the extraction steam from stages five to seven, condensate temperature of the lower header, fan speed feedback, fan current feedback, fan inverter operating status, fan inverter stopped status, exhaust isolation valve open status, fan reverse rotation status, and vacuum pump status.

[0011] Furthermore, the classification of freezing risk in each wind turbine area into three levels—low risk, medium risk, and high risk—based on the freezing probability includes:

[0012] When the probability of freezing is less than or equal to 0.4, it is classified as a low-risk level;

[0013] When the probability of freezing is greater than 0.4 and less than or equal to 0.7, it is classified as a medium-risk level;

[0014] When the probability of freezing is greater than 0.7, it is classified as a high-risk level.

[0015] Furthermore, the first optimization model takes the current fan speed, fan speed change, ambient temperature, wind speed, and wind direction as input features, and outputs the temperature change of the corresponding condenser temperature measurement zone. The first optimization model is established with the goal of minimizing the risk of freezing in the air-cooled island as shown below:

[0016] min Model risk (t condense ,t i,j +△t,t amb ...)

[0017] st△t i =Model i fan (△fan i )

[0018] Among them, Model risk This represents the air-cooled island freezing risk assessment model, where Δt represents the temperature change after a change in fan speed, and Δfan represents the temperature change after a change in fan speed. i t represents the change in wind speed of the i-th fan; condense Indicates the temperature of condensate; t amb Indicates ambient temperature; t i,j This represents the temperature of the measuring band in the i-th row and j-th column; Model i fan Denotes the wind speed model of the i-th wind turbine; △t i This represents the temperature change after the wind speed of the i-th fan changes.

[0019] The first optimization model is solved based on the set constraints, which include: total air volume is greater than or equal to minimum operating condensation demand, fans in high-risk areas cannot be shut down, and the maximum allowable number of fans to start and stop is less than or equal to the total number of fans.

[0020] Furthermore, the second optimization model is a multi-input multi-output model. First, models for the reverse fan speed, vacuum pump start / stop, and temperature measurement zone are established respectively. reverse and model pump Then, a second optimization model is established with the objective of minimizing the risk of freezing in the air-cooled island, as shown below:

[0021] min Model risk (t condense ,t i,j +△t,t amb ...)

[0022] t i,j＝ model j reverse +model j pump

[0023] The second optimization model is solved based on the set constraints, which include: the wind turbine frequency is less than the maximum wind turbine frequency, wind turbines in high-risk areas cannot be shut down, and the maximum allowable number of wind turbines to start and stop is less than or equal to the total number of wind turbines.

[0024] Furthermore, the automatic encoder model AE is trained using the root mean square error (MSE) as the loss function, and based on the error distribution, the upper 95th percentile is used as the alarm threshold.

[0025] Furthermore, when analyzing the collected air-cooled island data using the automatic encoder model AE, the surface temperature measurement zone of the air-cooled island with a value greater than 0 is set to 0.

[0026] This specification provides one or more embodiments of an air-cooled island anti-freezing control system based on AI prediction and two-stage optimization, including:

[0027] Risk assessment module: Used to analyze the collected air-cooled island data using the automatic encoder model AE, assess the freezing probability of each fan area within a given time window, and classify the freezing risk of each fan area into three levels: low risk, medium risk, and high risk based on the freezing probability.

[0028] The first optimization module is used to establish a first optimization model of the change in fan speed and the corresponding temperature change of the temperature measurement zone when the freezing risk is identified as medium or high, and to carry out the first stage of optimization with the goal of minimizing the freezing risk of the air-cooled island.

[0029] The second optimization module is used to establish a second optimization model for the reverse fan speed, vacuum pump start / stop and temperature measurement zone when the first optimization model cannot meet the antifreeze requirements. The second stage of optimization is aimed at minimizing the risk of freezing in the air-cooled island.

[0030] Control feedback module: Used to convert optimization results into control commands that can be issued, and to collect feedback signals to perform closed-loop correction of the automatic encoder model.

[0031] This specification provides one or more embodiments of an electronic device, including:

[0032] Processor; and,

[0033] A memory is configured to store computer-executable instructions, which, when executed, cause the processor to implement the steps of the aforementioned air-cooled island antifreeze control method based on AI prediction and two-stage optimization.

[0034] This specification provides one or more embodiments of a storage medium for storing computer-executable instructions that, when executed, implement the steps of the above-described air-cooled island antifreeze control method based on AI prediction and two-stage optimization.

[0035] By employing the embodiments of the present invention, through precise freezing risk assessment, a two-stage optimized control strategy, and a dynamic adjustment and closed-loop correction mechanism, energy-saving and efficient anti-freezing control of the air-cooled island is achieved. The freezing risk is accurately identified, and corresponding control measures are taken according to the risk level. The control strategy can be adjusted in real time according to the actual operating conditions to ensure stability and adaptability, improve the safety and reliability of the air-cooled island operation, reduce energy consumption, and have significant economic and social benefits.

[0036] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 A flowchart illustrating an air-cooled island antifreeze control method based on AI prediction and two-stage optimization, provided for one or more embodiments of this specification;

[0039] Figure 2 A flowchart illustrating the implementation of an air-cooled island antifreeze control method based on AI prediction and two-stage optimization, provided for one or more embodiments of this specification.

[0040] Figure 3 A schematic diagram illustrating the composition of an air-cooled island antifreeze control system based on AI prediction and two-stage optimization, provided for one or more embodiments of this specification;

[0041] Figure 4 This is a schematic diagram of the structure of an electronic device provided for one or more embodiments of this specification. Detailed Implementation

[0042] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.

[0043] Method Implementation Examples

[0044] According to embodiments of the present invention, an antifreeze control method for air-cooled islands based on AI prediction and two-stage optimization is provided. Figure 1 The flowchart illustrates an air-cooled island antifreeze control method based on AI prediction and two-stage optimization, provided for one or more embodiments of this specification. Figure 2A flowchart illustrating the implementation of an AI-based prediction and two-stage optimization-based antifreeze control method for air-cooled islands, provided for one or more embodiments of this specification, is shown below. Figure 1 and Figure 2 As shown, the air-cooled island antifreeze control method based on AI prediction and two-stage optimization according to an embodiment of the present invention specifically includes:

[0045] S1. The collected air-cooled island data is analyzed using the automatic encoder model AE to assess the freezing probability of each wind turbine area within a given time window, and the freezing risk of each wind turbine area is divided into three levels: low risk, medium risk, and high risk based on the freezing probability.

[0046] Data on the air-cooled island over the past year is acquired from the DCS system and edge sensors. This data includes: condensate temperature, surface temperature of the air-cooled island, ambient temperature, ambient humidity, fan speed, wind speed, wind direction, main steam flow, total coal consumption, total air volume, low-pressure cylinder inlet steam pressure, low-pressure cylinder inlet steam temperature, condensate pump outlet header temperature, temperature and pressure of steam extracted from stages five to seven, condensate temperature in the lower header, fan speed feedback, fan current feedback, fan inverter operating status, fan inverter stopped status, exhaust isolation valve open status, fan reverse rotation status, and vacuum pump status, excluding data related to icing.

[0047] Based on the acquired historical air-cooled island data, the automatic encoder model (AE) is used to predict the freezing probability of each wind turbine area within a given time window. According to the freezing probability, the freezing risk of each wind turbine area is divided into three levels: low risk, medium risk, and high risk. Specifically:

[0048] When the probability of freezing is less than or equal to 0.4, it is classified as a low-risk level;

[0049] When the probability of freezing is greater than 0.4 and less than or equal to 0.7, it is classified as a medium-risk level;

[0050] When the probability of freezing is greater than 0.7, it is classified as a high-risk level.

[0051] The autoencoder model is trained using the root mean square error (MSE) as the loss function. risk (t condense ,t i,j, t amb ...), as shown below:

[0052]

[0053] Where t i,j Identify the temperature of the temperature measurement zone in the i-th row and j-th column, and set the temperature measurement zone temperature of the air-cooled island surface that is greater than 0 to 0;

[0054] Based on the error distribution setting, the upper 95th percentile is used as the alarm threshold.

[0055] S2. When the freezing risk is identified as medium or high, a first optimization model is established to correlate the fan speed change with the corresponding temperature change of the temperature measurement zone. The first stage of optimization is carried out with the goal of minimizing the freezing risk of the air-cooled island.

[0056] When the freezing risk is identified as medium or high, the first stage of optimization for the air-cooled island freezing risk is performed by adjusting the fan frequency to reduce the freezing risk in the fan area. A first optimization model is established to correlate fan speed changes with corresponding temperature changes in the temperature measurement zone. The input features of the first optimization model are the current fan speed, fan speed changes, ambient temperature, wind speed, and wind direction. The output is the temperature change in the corresponding condenser temperature measurement zone. The first optimization model is established with the goal of minimizing the freezing risk of the air-cooled island, as shown below:

[0057] min Model risk (t condense ,t i,j +△t,t amb ...)

[0058] st△t i =Model i fan (△fan i )

[0059] Among them, Model risk This represents the air-cooled island freezing risk assessment model, where Δt represents the temperature change after a change in fan speed, and Δfan represents the temperature change after a change in fan speed. i t represents the change in wind speed of the i-th fan; condense Indicates the temperature of condensate; t amb Indicates ambient temperature; t i,j This represents the temperature of the measuring band in the i-th row and j-th column; Modell i fan Denotes the wind speed model of the i-th wind turbine; △t i This represents the temperature change after the wind speed of the i-th fan changes.

[0060] The first optimization model is solved according to the set constraints, which include: the total air volume is greater than or equal to the minimum operating condensation requirement, the fans in high-risk areas cannot be shut down, and the maximum allowable number of fans to start and stop is less than or equal to the total number of fans.

[0061] S3. When the first optimization model cannot meet the antifreeze requirements, a second optimization model is established, which includes the reverse fan speed, vacuum pump start / stop, and temperature measurement zone. The second stage of optimization is carried out with the goal of minimizing the risk of freezing in the air-cooled island.

[0062] When fan control cannot fully meet antifreeze requirements, for high-risk areas where the first-stage control fails, reversing the fan and vacuum pump can further reduce the risk of icing in the air-cooled island. Based on the relationship between reversing fan speed, vacuum pump start / stop, and temperature measurement zone, a second optimization model is established. Since the reversing fan and vacuum pump affect the temperature measurement zone of the entire column, the second optimization model is a multi-input multi-output model, using weighted optimization adjustment of the reversing fan and vacuum pump, with the reversing fan control having higher priority than the vacuum pump. First, models for reversing fan speed, vacuum pump start / stop, and temperature measurement zone are established respectively. reverse and model pump Among them, the model of the reverse fan speed and the temperature of the temperature measuring zone. reverse The input features are the current wind speed of the i-th column reverse fan, the wind speed change of the i-th column reverse fan, the ambient temperature, wind speed, and wind direction. The model also includes the vacuum pump start / stop and the temperature measurement zone. pump The input features are the start / stop of the vacuum pump in the i-th column, ambient temperature, wind speed, and wind direction. Then, a second optimization model is established with the objective of minimizing the freezing risk of the air-cooled island, as shown below:

[0063] min Model risk (t condense ,t i,j +△t,t amb ...)

[0064] t i,j＝ model j reverse +model j pump

[0065] The second optimization model is solved based on the set constraints, which include: the wind turbine frequency is less than the maximum wind turbine frequency, wind turbines in high-risk areas cannot be shut down, and the maximum allowable number of wind turbines to start and stop is less than or equal to the total number of wind turbines.

[0066] S4. Convert the optimization results into control commands that can be issued, and collect feedback signals to perform closed-loop correction of the automatic encoder model.

[0067] The beneficial effects of this invention are as follows:

[0068] This invention, through error distribution quantile modeling and setting dynamic risk levels (low, medium, and high), can more accurately reflect the actual icing trend, facilitating graded response and progressive control. Through precise freezing risk assessment, a two-stage optimized control strategy, and a dynamic adjustment and closed-loop correction mechanism, it achieves energy-saving and efficient anti-freezing control for air-cooled islands. It accurately identifies freezing risks and takes corresponding control measures based on the risk level. The control strategy can be adjusted in real time according to actual operating conditions, ensuring stability and adaptability, improving the safety and reliability of air-cooled island operation, and reducing energy consumption, resulting in significant economic and social benefits.

[0069] System Implementation Examples

[0070] According to embodiments of the present invention, an anti-freezing control system for air-cooled islands based on AI prediction and two-stage optimization is provided. Figure 3 This specification provides a schematic diagram of the composition of an air-cooled island anti-freezing control system based on AI prediction and two-stage optimization, for one or more embodiments. Figure 3 As shown, the air-cooled island antifreeze control system based on AI prediction and two-stage optimization according to an embodiment of the present invention specifically includes:

[0071] Risk assessment module 30: Used to analyze the collected air-cooled island data using the automatic encoder model AE, assess the freezing probability of each fan area within a given time window, and classify the freezing risk of each fan area into three levels: low risk, medium risk, and high risk based on the freezing probability.

[0072] First optimization module 32: When the freezing risk is identified as medium or high, it establishes a first optimization model of the fan speed change and the corresponding temperature change of the temperature measurement zone, and performs the first stage optimization with the goal of minimizing the freezing risk of the air-cooled island.

[0073] Second optimization module 34: When the first optimization model cannot meet the antifreeze requirements, a second optimization model is established for the reverse fan speed, vacuum pump start / stop and temperature measurement zone temperature, with the goal of minimizing the freezing risk of the air-cooled island;

[0074] Control feedback module 36: used to convert optimization results into control commands that can be issued, and to collect feedback signals to correct the automatic encoder model in a closed loop.

[0075] The embodiments of the present invention are system embodiments corresponding to the above method embodiments. The specific operation of each module can be understood by referring to the description of the method embodiments, and will not be repeated here.

[0076] Device Example 1

[0077] This invention provides an electronic device, such as... Figure 4As shown, it includes: a memory 40, a processor 42, and a computer program stored in the memory 40 and executable on the processor 42. When the computer program is executed by the processor 42, it performs the following method steps:

[0078] S1. The collected air-cooled island data is analyzed using the automatic encoder model AE to assess the freezing probability of each wind turbine area within a given time window, and the freezing risk of each wind turbine area is divided into three levels: low risk, medium risk and high risk based on the freezing probability.

[0079] S2. When the freezing risk is identified as medium or high, a first optimization model is established to correlate the fan speed change with the corresponding temperature change of the temperature measurement zone. The first stage of optimization is carried out with the goal of minimizing the freezing risk of the air-cooled island.

[0080] S3. When the first optimization model cannot meet the antifreeze requirements, a second optimization model is established, which includes the reverse fan speed, vacuum pump start / stop and temperature measurement zone temperature. The second stage of optimization is carried out with the goal of minimizing the risk of freezing in the air-cooled island.

[0081] S4. Convert the optimization results into control commands that can be issued, and collect feedback signals to perform closed-loop correction of the automatic encoder model.

[0082] Device Example 2

[0083] This invention provides a computer-readable storage medium storing an information transmission implementation program. When executed by a processor 42, the program performs the following method steps:

[0084] S1. The collected air-cooled island data is analyzed using the automatic encoder model AE to assess the freezing probability of each wind turbine area within a given time window, and the freezing risk of each wind turbine area is divided into three levels: low risk, medium risk and high risk based on the freezing probability.

[0085] S2. When the freezing risk is identified as medium or high, a first optimization model is established to correlate the fan speed change with the corresponding temperature change of the temperature measurement zone. The first stage of optimization is carried out with the goal of minimizing the freezing risk of the air-cooled island.

[0086] S3. When the first optimization model cannot meet the antifreeze requirements, a second optimization model is established, which includes the reverse fan speed, vacuum pump start / stop and temperature measurement zone temperature. The second stage of optimization is carried out with the goal of minimizing the risk of freezing in the air-cooled island.

[0087] S4. Convert the optimization results into control commands that can be issued, and collect feedback signals to perform closed-loop correction of the automatic encoder model.

[0088] The computer-readable storage media described in this embodiment include, but are not limited to, ROM, RAM, disk, or optical disk.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for antifreezing control of air-cooled islands based on AI prediction and two-stage optimization, characterized in that, include: The collected air-cooled island data is analyzed using the automatic encoder model AE to assess the freezing probability of each wind turbine area within a given time window, and the freezing risk of each wind turbine area is divided into three levels: low risk, medium risk, and high risk based on the freezing probability. When the freezing risk is identified as medium or high, a first optimization model is established to correlate the fan speed change with the corresponding temperature change of the temperature measurement zone. The first stage of optimization is carried out with the goal of minimizing the freezing risk of the air-cooled island. When the first optimization model cannot meet the antifreeze requirements, a second optimization model is established, which includes the reverse fan speed, vacuum pump start / stop and temperature measurement zone temperature. The second stage of optimization is carried out with the goal of minimizing the risk of freezing in the air-cooled island. The optimization results are converted into control commands that can be issued, and feedback signals are collected to perform closed-loop correction of the automatic encoder model.

2. The method according to claim 1, characterized in that, Data on the air-cooled island is acquired from the DCS system and edge sensors. This data includes: condensate temperature, surface temperature of the air-cooled island, ambient temperature, ambient humidity, fan speed, wind speed, wind direction, main steam flow rate, total coal consumption, total air volume, low-pressure cylinder inlet pressure, low-pressure cylinder inlet temperature, condensate pump outlet header temperature, temperature and pressure of the extraction steam from stages five to seven, condensate temperature in the lower header, fan speed feedback, fan current feedback, fan inverter operating status, fan inverter stopped status, exhaust isolation valve open status, fan reverse rotation status, and vacuum pump status.

3. The method according to claim 1, characterized in that, The classification of freezing risk in each wind turbine area into three levels—low risk, medium risk, and high risk—based on freezing probability includes: When the probability of freezing is less than or equal to 0.4, it is classified as a low-risk level; When the probability of freezing is greater than 0.4 and less than or equal to 0.7, it is classified as a medium-risk level; When the probability of freezing is greater than 0.7, it is classified as a high-risk level.

4. The method according to claim 1, characterized in that, The first optimization model takes the current fan speed, fan speed change, ambient temperature, wind speed, and wind direction as input features, and outputs the temperature change of the corresponding condenser temperature measurement zone. The first optimization model is established with the goal of minimizing the risk of freezing in the air-cooled island as shown below: min Model risk (t condense ,t i,j +△t,t amb ...) s.t.△t i =Model i fan (△fan i ) Among them, Model risk This represents the air-cooled island freezing risk assessment model, where Δt represents the temperature change after a change in fan speed, and Δfan... i t represents the change in wind speed of the i-th fan; condense Indicates the temperature of condensate; t amb Indicates ambient temperature; t i,j This represents the temperature of the measuring band in the i-th row and j-th column; Model i fan Denotes the wind speed model of the i-th wind turbine; △t i This represents the temperature change after the wind speed of the i-th fan changes. The first optimization model is solved based on the set constraints, which include: total air volume is greater than or equal to minimum operating condensation demand, fans in high-risk areas cannot be shut down, and the maximum allowable number of fans to start and stop is less than or equal to the total number of fans.

5. The method according to claim 1, characterized in that, The second optimization model is a multi-input multi-output model. First, models for the reverse fan speed, vacuum pump start / stop, and temperature measurement zone are established respectively. reverse and model pump Then, a second optimization model is established with the objective of minimizing the risk of freezing in the air-cooled island, as shown below: min Model risk (t condense ,t i,j +△t,t amb ...) t i,j＝ model j reverse +model j pump The second optimization model is solved based on the set constraints, which include: the wind turbine frequency is less than the maximum wind turbine frequency, wind turbines in high-risk areas cannot be shut down, and the maximum allowable number of wind turbines to start and stop is less than or equal to the total number of wind turbines.

6. The method according to claim 1, characterized in that, The automatic encoder model AE is trained using the root mean square error (MSE) as the loss function and the upper 95th percentile is used as the alarm threshold based on the error distribution.

7. The method according to claim 2, characterized in that, When analyzing the collected air-cooled island data using the automatic encoder model AE, the surface temperature measurement zone of the air-cooled island with a value greater than 0 is set to 0.

8. A freeze-proof control system for air-cooled islands based on AI prediction and two-stage optimization, characterized in that, include: Risk assessment module: Used to analyze the collected air-cooled island data using the automatic encoder model AE, assess the freezing probability of each fan area within a given time window, and classify the freezing risk of each fan area into three levels: low risk, medium risk, and high risk based on the freezing probability. The first optimization module is used to establish a first optimization model of the change in fan speed and the corresponding temperature change of the temperature measurement zone when the freezing risk is identified as medium or high, and to carry out the first stage of optimization with the goal of minimizing the freezing risk of the air-cooled island. The second optimization module is used to establish a second optimization model for the reverse fan speed, vacuum pump start / stop and temperature measurement zone when the first optimization model cannot meet the antifreeze requirements. The second stage of optimization is aimed at minimizing the risk of freezing in the air-cooled island. Control feedback module: Used to convert optimization results into control commands that can be issued, and to collect feedback signals to perform closed-loop correction of the automatic encoder model.

9. An electronic device, characterized in that, include: processor; as well as, A memory configured to store computer-executable instructions, which, when executed, cause the processor to implement the steps of the air-cooled island antifreeze control method based on AI prediction and two-stage optimization as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, Used to store computer-executable instructions, which, when executed, implement the steps of the air-cooled island antifreeze control method based on AI prediction and two-stage optimization as described in any one of claims 1 to 7.

Images